Paper deep dive
Technological Advances in Detecting and Managing Cognitive Impairment in Older Adults: Trends, Challenges, and Future Directions
Mohammad Asif, Azizuddin Khan, Mohd Azam, Anurag Rajkumar Bombarde
Intelligence
Status: succeeded | Model: Gemma-4-26B-A4B | Prompt: intel-v1 | Confidence: 92%
Last extracted: 8/3/2026, 2:03:14 AM
Summary
This review synthesizes technological advances in detecting and managing cognitive impairment in older adults, focusing on neurophysiological signals (EEG), neuroimaging (MRI, PET), blood-based biomarkers, and digital markers integrated with AI/ML. It highlights the potential of deep learning models and multimodal fusion for early detection while emphasizing challenges in validation, standardization, and clinical translation. Key milestones include the approval of anti-amyloid therapies and blood-based tests for Alzheimer's disease.
Entities (19)
Relation Signals (14)
Plasma p-tau217 → aidsdiagnosisof → Alzheimer's Disease
confidence 96% · plasma p-tau217 has reached clinical utility, with the first blood test cleared to aid Alzheimer’s diagnosis
Mild Cognitive Impairment → isprecursorto → Alzheimer's Disease
confidence 95% · cognitive decline along the continuum from mild cognitive impairment (MCI) to dementia
Donanemab → treats → Alzheimer's Disease
confidence 94% · anti-amyloid therapies (lecanemab, donanemab) are approved
Lecanemab → treats → Alzheimer's Disease
confidence 94% · anti-amyloid therapies (lecanemab, donanemab) are approved
P300 → ismarkerin → Electroencephalography
confidence 93% · EEG markers (alpha/theta changes, P300 latency)
Electroencephalography → detects → Mild Cognitive Impairment
confidence 92% · EEG provides non-invasive, high-temporal-resolution insights into neural oscillations... robust early biomarkers of MCI
Theta Oscillations → ismarkerin → Electroencephalography
confidence 92% · EEG markers (alpha/theta changes, P300 latency)
Alpha Oscillations → ismarkerin → Electroencephalography
confidence 92% · EEG markers (alpha/theta changes, P300 latency)
Cypher Suggestions (0)
No Cypher suggestions yet.
Abstract
Abstract:As populations age, cognitive decline from mild cognitive impairment (MCI) to dementia is a defining health challenge of the coming decades, yet routine assessment often misses its earliest signs. This article critically synthesizes recent technological advances for detecting and managing cognitive impairment in older adults, spanning neurophysiological signals (chiefly electroencephalography, EEG), structural and molecular neuroimaging (MRI and amyloid/tau PET), blood-based biomarkers, and digital markers, integrated through artificial intelligence (AI), machine learning (ML), and deep learning (DL). Beyond summarizing, it contributes a cross-disciplinary taxonomy, a methodological-rigor lens foregrounding subject- and site-independent validation, an integrative early-detection framework linking tiered screening to intervention, and comparison tables of detection methods, interventions, and risk and protective factors. EEG markers (alpha/theta changes, P300 latency) and deep models (CNNs, LSTM/BiLSTM, transformers, self-supervised EEG foundation models) report strong accuracy, yet many rest on small, single-site datasets unlikely to survive rigorous external validation. Elsewhere, gains are tangible: plasma p-tau217 has reached clinical utility, with the first blood test cleared to aid Alzheimer's diagnosis in 2025; anti-amyloid therapies (lecanemab, donanemab) are approved despite modest, contested benefits; and multidomain lifestyle prevention has matured. Wearable, remote, speech, and virtual-reality tools enable continuous, ecologically valid monitoring, and multimodal fusion improves sensitivity and specificity. Barriers remain: standardization, explainability, data privacy, and equitable, externally validated deployment. The field's near-term promise lies in trustworthy, multimodal, longitudinally validated systems linking early detection to actionable, personalized care.
Tags
Links
- Source: https://arxiv.org/abs/2607.28687v1
- Canonical: https://arxiv.org/abs/2607.28687v1
Trouble viewing inline? Open PDF directly →
Full Text
152,370 characters extracted from source content.
Expand or collapse full text
Date of publication x 00, 0000, date of current version x 00, 0000. Digital Object Identifier 10.1109/ACCESS.2026.DOINUMBER Technological Advances in Detecting and Managing Cognitive Impairment in Older Adults: Trends, Challenges, and Future Directions MOHAMMAD ASIF 12 , AZIZUDDIN KHAN 1 , MOHD AZAM 2 , and ANURAG RAJKUMAR BOMBARDE 2 1 Indian Institute of Technology Bombay, Mumbai 400076, India (e-mail: pse2017001@gmail.com; khanaziz@iitb.ac.in; ORCID: 0000-0002-9517-6716 (Mohammad Asif)) 2 T-Systems ICT India Pvt. Ltd., Pune, India (e-mail: pse2017001@gmail.com; azam.251181@gmail.com; anurag.bombarde.dev@gmail.com) Corresponding author: Mohammad Asif (e-mail: pse2017001@gmail.com). ABSTRACT As populations age, cognitive decline along the continuum from mild cognitive impairment (MCI) to dementia is becoming one of the defining health challenges of the coming decades, yet routine clinical assessment still tends to miss its earliest and subtlest signs. This article surveys and critically synthesizes recent technological advances for detecting and managing cognitive impairment in older adults, drawing together neurophysiological signals (chiefly electroencephalography, EEG), structural and molecular neuroimaging (MRI and amyloid/tau PET), blood-based biomarkers, and digital markers, together with their integration through artificial intelligence (AI), machine learning (ML), and deep learning (DL). The review does more than summarize this literature: it contributes a unified cross- disciplinary taxonomy, a methodological-rigor lens that keeps subject- and site-independent validation in view, an integrative early-detection framework connecting tiered screening to timely intervention, and comparison tables of detection methods, interventions, and risk and protective factors. Across modalities, EEG markers (alpha and theta changes, P300 latency) and deep models (CNNs, LSTM/BiLSTM, transformers, and self-supervised EEG foundation models) report strong diagnostic accuracy, yet we argue that many such figures rest on small, single-site datasets unlikely to hold up under rigorous external validation. Elsewhere the gains are tangible: plasma phosphorylated tau-217 has reached clinical utility, with the first blood test cleared to aid Alzheimer’s diagnosis in 2025; anti-amyloid therapies (lecanemab, donanemab) have been approved despite modest, contested benefits; and multidomain lifestyle prevention has matured, targeting a large and potentially modifiable share of dementia risk. Wearable, remote, speech and large-language-model, and virtual-reality tools now support continuous, ecologically valid monitoring, and multimodal fusion tends to improve sensitivity and specificity. Substantial barriers remain, among them standardization, explainability, data privacy, and equitable, externally validated deployment via approaches such as federated learning. The field’s near-term promise, we conclude, lies in trustworthy, multimodal, longitudinally validated systems that link early detection to actionable, personalized care. INDEX TERMS Cognitive impairment, mild cognitive impairment, dementia, Alzheimer’s disease, electroencephalography (EEG), magnetic resonance imaging (MRI), machine learning, deep learning, foundation models, digital biomarkers, blood-based biomarkers, wearable sensing, review. I. INTRODUCTION T HE world is aging. As life expectancy rises across most regions and populations grow older, age-related cognitive disorders, particularly Mild Cognitive Impairment (MCI) and dementia, are becoming more common, placing a growing burden on individuals, families, and healthcare systems worldwide. Historically, the diagnosis and management of cognitive decline have relied on in-clinic assessments that are time- consuming, resource-intensive, and often insensitive to the earliest and most treatable changes. Recent technological developments are beginning to address these limitations. VOLUME 11, 20231 arXiv:2607.28687v1 [cs.LG] 30 Jul 2026 2023202420252026 Lecanemab: FDA traditional approval [6] Aducanumab discontinued [7] Donanemab approval; revised A criteria adopt blood biomarkers [8], [9] Lancet Commission: 14 modifiable risk factors [10] EEG foundation models (LaBraM, CBraMod) [11], [12] FDA clears first blood test (p-tau217/Aβ42) [13] US POINTER: multidomain prevention at scale [14] FIGURE 1. Selected milestones (2023–2026) in the detection, treatment, and prevention of cognitive impairment, spanning disease-modifying therapies, blood-based biomarkers, updated diagnostic criteria, large-scale prevention trials, and self-supervised EEG foundation models. Advances in brain imaging, notably EEG and functional MRI, are increasingly paired with artificial intelligence, machine learning, and deep learning to produce tools that are more precise, accessible, and scalable. Together, they open a path from reactive, one-size-fits-all care toward proactive and personalized approaches grounded in real-world context. This shift enables earlier detection, more timely support, and better outcomes for aging populations. Population aging has sharpened attention on cognitive impairment in older adults, where MCI affects a substan- tial share of this group and often precedes dementia and Alzheimer’s disease [1], [2]. Recent advances in electroen- cephalography (EEG), artificial intelligence (AI), machine learning (ML), and deep learning (DL) have opened new op- portunities for the early detection, assessment, and interven- tion of cognitive decline in this population [3], [4], building on foundational work in computational neuroscience that spans neural modeling, deep learning, and the encoding and decoding of neural signals [5]. Yet despite this rapid proliferation of new technologies, the evidence remains fragmented and scattered across dis- ciplines. A structured synthesis is needed to consolidate what is known. This article therefore reviews the breadth and nature of the literature on these technological advances for detecting and managing MCI and dementia in older adults. Figure 1 highlights key milestones from 2023– 2026 that illustrate the field’s rapid recent progress across therapeutics, biomarkers, diagnostic criteria, prevention, and modeling. A. BACKGROUND AND RATIONALE The global demographic shift toward an aging population has led to a sharp rise in the prevalence of cognitive disorders, with over 55 million people currently living with dementia worldwide and nearly 10 million new cases each year [10]. Mild cognitive impairment (MCI), a prodromal stage between normal aging and dementia, affects approx- imately 19.7–23.7% of older adults, varying by setting and assessment tool [15], [16]. Throughout this article we distinguish subjective cognitive decline (SCD), amnestic and non-amnestic MCI, and dementia along this clinical continuum, and refer to the amyloid/tau/neurodegeneration (A/T/N) biomarker framework where relevant. Traditional paper-and-pencil assessments, such as the Mini-Mental State Examination (MMSE), suffer from ceiling effects, cultural bias, and limited sensitivity for early changes, resulting in delayed detection of subtle declines [17]. These limitations, combined with resource constraints in clinical and commu- nity settings, underscore the need for objective, scalable, and continuous monitoring solutions. Recent advances in neurotechnology, artificial intelli- gence (AI), and digital health have opened new possibilities for the early detection and personalized management of cognitive impairment. Electroencephalography (EEG) pro- vides non-invasive, high-temporal-resolution insights into neural oscillations and event-related potentials, with alpha power reductions and P300 latency increases serving as robust early biomarkers of MCI and Alzheimer’s disease (AD) [18], [19]. Deep learning models, particularly con- volutional neural networks (CNNs) and long short-term memory (LSTM) networks, achieve diagnostic accuracies exceeding 95% by automatically extracting complex spa- tiotemporal features from EEG and neuroimaging data [2], [20], [21]. Multimodal approaches integrating EEG, MRI radiomics, and wearable sensors further enhance sensitivity and specificity, enabling comprehensive characterization of both functional and structural brain changes [3], [22]. Wearable EEG devices and smartphone-based digital phe- notyping platforms support continuous, real-world moni- toring of cognitive function, capturing naturalistic brain activity, gait, and behavioral markers outside the clinic [23], [24]. Virtual reality (VR) and immersive assessments simulate everyday tasks, uncovering subtle deficits that stan- dard tests miss [25], [26]. Non-invasive brain stimulation (tDCS, TMS) and adaptive neurofeedback use real-time EEG feedback for personalized cognitive enhancement [27], [28]. Despite these promising developments, challenges re- main in standardizing protocols, ensuring data privacy, and achieving regulatory approval for AI-driven medical devices. Large-scale, multicenter validation studies and consensus guidelines are essential to translate these innovations into routine care. This review maps the technological advances in cognitive impairment detection and management, identifies research gaps, and outlines future directions for precision cognitive healthcare. B. OBJECTIVES This review sets out to map and synthesize current techno- logical advances for the early detection and management 2VOLUME 11, 2023 of cognitive impairment in older adults, with particular attention to how these innovations intersect with lifestyle factors and the progression to long-term impairments [29], [30]. Specifically, this review seeks to address the following research questions: • What neurophysiological technologies (e.g., EEG, neu- roimaging, digital biomarkers) and computational tools (AI, ML, DL, NLP) are currently used for the early detection of cognitive decline in older adults, and how do they compare in accuracy and feasibility [29], [31], [32]? • How are digital and wearable technologies, including contactless and remote monitoring approaches, being used for continuous assessment and early intervention [33], [34]? • What is the evidence for the influence of lifestyle factors—such as physical activity, diet, social engage- ment, sleep, and psychosocial wellbeing—on cognitive trajectories, and how are these factors integrated into technological risk models or intervention strategies [35]–[38]? • Which risk and protective factors are most strongly associated with the progression from early cognitive decline or MCI to long-term dementia, and how can early detection facilitate timely preventive or therapeu- tic interventions [30], [31], [39], [40]? • What are the major barriers and facilitators to clinical translation, implementation, and real-world adoption of these technological solutions in diverse populations and care settings [33], [41]? • Where are the key evidence gaps, and what future research directions are needed to optimize early detec- tion, address modifiable lifestyle factors, and ultimately reduce the burden of dementia? By addressing these questions, the review offers an inte- grative account of how emerging technologies and lifestyle considerations can be combined to improve the early detec- tion, risk stratification, and management of cognitive im- pairment, ultimately aiming to delay or prevent progression to dementia in aging populations. C. SCOPE AND ARTICLE SELECTION This article is a narrative (critical) survey rather than a systematic or scoping review; accordingly, it does not re- port formal screening counts or a PRISMA flow diagram, and its coverage is intended to be representative of the major trends rather than exhaustive. Relevant literature was identified through the principal bibliographic and full-text databases spanning the biomedical and computing disci- plines (including PubMed/PMC, IEEE Xplore, the ACM Digital Library, Scopus, Web of Science, ScienceDirect, and SpringerLink), supplemented by discovery and metadata services (Google Scholar, Semantic Scholar, and Crossref) for forward/backward citation tracing and bibliographic ver- ification. Search terms combined concepts of the target pop- ulation (older adults, elderly, aging), the clinical condition (mild cognitive impairment, MCI, dementia, Alzheimer’s disease, cognitive decline), and the enabling technologies (EEG, MRI/fMRI, artificial intelligence, machine learning, deep learning, transformer, foundation model, wearable, digital biomarker, speech/language, virtual reality, blood- based biomarker). Emphasis was placed on peer-reviewed work published between 2020 and 2026 to capture the current state of the art, while seminal earlier studies were retained where they anchor foundational concepts. Prior- ity was given to articles in indexed, peer-reviewed jour- nals (SCI/SCIE/Scopus) and top-tier peer-reviewed con- ferences; preprints were consulted only where no peer- reviewed equivalent was available and are identified as such. Studies were considered relevant when they addressed the detection, monitoring, risk stratification, or management of cognitive impairment in older adults using one or more of these technologies, with a deliberate effort to balance clin- ical/biomedical and computational (AI/ML/DL and signal- processing) perspectives. D. CONTRIBUTIONS AND ORGANIZATION OF THIS REVIEW The main contributions of this review are fourfold. First, we present a unified taxonomy (Fig. 2) that spans neuro- physiological, molecular, digital, and lifestyle evidence for the detection and management of cognitive impairment in older adults. Second, we apply a consistent methodological- rigour lens throughout, foregrounding subject- and site- independent external validation, to temper the headline accuracy figures that pervade this literature. Third, we propose an integrative early-detection framework that links complementary screening modalities to timely intervention. Fourth, we consolidate the evidence into comparison tables of detection methods, interventions, and risk and protective factors. In contrast to recent modality-specific surveys (for example on EEG [42], multimodal fusion [43], speech and language [44], wearables [24], virtual reality [45], and remote assessment [46]), this article integrates these strands into a single cross-disciplinary account oriented toward early, trustworthy, and clinically deployable detection. The remainder of this article is organized as follows. We first survey detection technologies by modality: EEG- and MRI-based neurophysiological approaches; AI, ML, and deep-learning methods, including transformers and EEG foundation models; digital, speech/language, virtual-reality, and wearable biomarkers; and advanced signal-processing, portable-sensing, and clinical-validation techniques. We then turn to prevention and early detection by integrating mod- ifiable lifestyle factors, presenting an integrative early- detection framework, and synthesizing the risk and protec- tive factors that govern progression from MCI to dementia. Finally, we discuss clinical-translation and implementation challenges, outline future research directions, and state the limitations of this review before concluding. VOLUME 11, 20233 Technologies for early detection & management of cognitive impairment AI / ML methodsDetection modalities Management & prevention EEG & event- related potentials MRI / PET neuroimaging Blood biomarkers (p- tau217, Aβ42/40) Digital: speech/LLM, wearables, VR Classical ML (SVM, TDA) CNN / RNN (LSTM, BiLSTM) Transformers & foundation models Multimodal fusion Lifestyle / multidomain prevention Pharmacological (lecanemab, donanemab) Neuromodulation (40-Hz tACS) Cross-cutting requirements: explainable AI· federated / privacy- preserving learning· standardization & external validation FIGURE 2. A taxonomy of technologies for the early detection and management of cognitive impairment in older adults, organized by detection modality, AI/ML method family, and management/prevention approach, together with cross-cutting requirements that condition clinical translation. I. EEG AND MRI-BASED APPROACHES FOR COGNITIVE ASSESSMENT A. NEUROPHYSIOLOGICAL BIOMARKERS IN COGNITIVE DECLINE EEG and MRI/fMRI are powerful tools for investigating cognitive impairment because they measure neural activity directly and non-invasively, with precise temporal or spatial resolution [47], [48]. Recent EEG work shows real potential for early detection of cognitive decline through the analy- sis of brain oscillations, connectivity patterns, and event- related potentials. Techniques such as frequency-band and connectivity analysis illuminate distinct aspects of cognitive function [48]. Distinct oscillatory and connectivity changes across the classical EEG frequency bands, together with event-related potential alterations, characterize cognitive de- cline, and combining several such biomarkers generally outperforms single-parameter approaches [49], [50]. Studies consistently show that older adults with cogni- tive impairment exhibit distinct EEG patterns relative to cognitively healthy controls [48], [51]. Alpha oscillations have drawn particular attention as candidate biomarkers of cognitive decline [18], [52]. Alpha power and functional connectivity differ significantly between cognitively intact older adults and those with MCI or Alzheimer’s disease [18], [53]. A systematic review of EEG connectivity in MCI and AD found an overall pattern of lower connectivity than in healthy controls, most prominent in the alpha band [18]. The clinical meaning of alpha-rhythm changes goes beyond simple power measures to encompass connectivity patterns that reflect the integrity of the neural networks underlying cognition. Event-related potentials (ERPs), particularly the P300 component, correlate with cognitive function in older adults, and P300 amplitude and latency can distinguish healthy aging from pathological cognitive decline [4], [47], [54]. Some studies report P300 abnormalities in prodromal and MCI stages, suggesting utility for early intervention, although evidence in biomarker-defined preclinical AD re- mains limited [19]. MRI is a cornerstone of neuroimaging, and its integration with artificial intelligence (AI) is now extending what it can reveal. This review synthesizes progress in coupling MRI with convolutional neural networks (CNNs), not only to detect cognitive impairment with high accuracy but also to confront the harder problems of model generalizability and clinical transparency. CNN-based models can classify stages of dementia with very high accuracy: one deep- learning pipeline analyzing 3D MRI data reached 99.94% accuracy in distinguishing Non-dementia (ND), Very Mild Dementia (VMD), Mild Dementia (MD), and Moderate Dementia (MoD) [55]; as with many single-dataset fig- ures reported here, such accuracy should be interpreted cautiously pending subject- and site-independent external validation (cf. Table 6). Adding multi-modal imaging such as diffusion-weighted MRI (dMRI) alongside standard T1- weighted scans improves diagnostic performance [56]. To counter dataset bias and improve performance across pop- ulations (e.g., North American vs. Indian cohorts), 3D Cy- cleGAN harmonization standardizes imaging data and yields better classification [56]. More recent methods move beyond CycleGANs toward conditional latent-diffusion models, and open benchmarking toolkits now allow multi-site harmo- nization approaches to be compared on a common footing [57], [58]. Applied to MRI, explainable AI supports feature- attribution methods such as Layer-wise Relevance Propaga- tion (LRP), which produce visual heatmaps of diagnostically relevant brain regions [59]. 4VOLUME 11, 2023 B. FREQUENCY BAND ANALYSIS AND COGNITIVE FUNCTION Different EEG frequency bands reveal different facets of cognition. Gamma oscillations, which support cognitive binding and attention, show reduced power in both healthy aging and mild cognitive impairment (MCI) [48]. This re- duction is thought to reflect impaired neural synchronization and network integration, processes essential for higher-order cognition [48], [51]. Theta band activity is closely associated with memory processes, particularly encoding and retrieval, and exhibits characteristic alterations in cognitive impairment [48]. In- creased theta power and disrupted theta connectivity have been reported in individuals with MCI relative to healthy controls, though findings vary across studies, and are in- terpreted as compensatory or pathological changes in hip- pocampal and prefrontal networks [53], [60]. A recent study examining functional connectivity patterns found distinct differences across age and MCI groups, with healthy aging showing decreased alpha functional connec- tivity while MCI specifically reduced gamma connectivity compared to age-matched controls [48]. These findings sug- gest different mechanisms underlying normal aging versus pathological cognitive decline. Combining multiple EEG biomarkers (power and con- nectivity across several frequency bands) improves discrim- ination among healthy aging, MCI, and early Alzheimer’s disease relative to single-parameter approaches, raising sen- sitivity and specificity and supporting its use for early detection and longitudinal monitoring of cognitive decline [51], [53], [61]. Recent studies reinforce and extend these directions: the resting-state theta/alpha power ratio discrim- inates amnestic MCI [62], sleep-like slow waves during wakefulness track amyloid burden and neurodegeneration in preclinical Alzheimer’s disease [63], and task-based paradigms are increasingly used to elicit condition-specific EEG markers [64]; on the management side, EEG-informed 40-Hz (gamma) stimulation is being trialed to modulate these oscillatory markers [65]. C. STRUCTURAL, FUNCTIONAL, AND MOLECULAR NEUROIMAGING Beyond electrophysiology, structural, functional, and molec- ular neuroimaging offer complementary windows on neu- rodegeneration. On structural MRI, medial-temporal and hippocampal atrophy together with regional cortical thinning are among the most reproducible correlates of progression from MCI to dementia [66], [67]. Diffusion MRI adds sensitivity to white-matter microstructural change, and com- bining diffusion-weighted with T1-weighted sequences has been used to improve automated staging [56]. Functional MRI captures the breakdown of large-scale networks, no- tably altered default-mode-network connectivity, that ac- companies early Alzheimer’s disease and cerebral small- vessel pathology [68], [69]. Molecular imaging with amy- loid and tau positron-emission tomography (PET) visual- izes the underlying proteinopathy directly and, with fluid biomarkers, anchors the biological staging formalized in the 2024 diagnostic criteria [9]. Because these acquisitions vary across scanners and sites, cross-scanner harmonization (increasingly via conditional latent-diffusion methods) and open benchmarking are prerequisites for generalizable imag- ing markers [57], [58]. Increasingly, these modalities are combined in machine-learning prediction models: pipelines that integrate structural MRI, amyloid PET, and clinical measures, or that apply automated analysis to cerebral magnetic resonance angiography, have been used to predict cognitive impairment in non-demented older adults [70], [71]. I. ARTIFICIAL INTELLIGENCE AND MACHINE LEARNING APPLICATIONS Artificial intelligence and machine learning are now central to efforts to predict and detect cognitive decline in older adults, spanning classical models, deep networks, and mul- timodal data fusion [72]. A. DEEP LEARNING ARCHITECTURES FOR COGNITIVE ASSESSMENT Deep learning has been notably successful at analyzing complex EEG patterns for cognitive impairment detection [2], [4]. Long Short-Term Memory (LSTM) networks are particularly well suited to EEG because they capture tem- poral dependencies in neural signals [2], [73]. One study built an LSTM-based framework for early MCI detection from EEG. Its pipeline moved through denoising to remove artifacts, segmentation and downsampling to stan- dardize the data, and deep feature extraction with stacked LSTM layers, letting the model surface subtle biomarkers in the EEG time series that conventional methods typically miss. Under cross-validation on a publicly available MCI EEG dataset, the best LSTM model reached 96.41% accu- racy, 96.55% sensitivity, and 95.95% specificity, pointing to its potential for automated screening and diagnosis of cognitive impairment [2]. Bidirectional LSTM (BiLSTM) networks improve pre- diction further by processing EEG in both temporal di- rections, capturing richer contextual information. Recent BiLSTM models, often combined with convolutional layers or attention mechanisms, achieve accuracies exceeding 96% for multi-class classification of Alzheimer’s disease, MCI, and healthy controls; reported figures include 96.03% and 97.31% on benchmark EEG datasets and 96.52% for a convolutional–attention–BiLSTM fusion model [4], [21], [74]. Such architectures retain spatial relationships in ERP data while learning long-term temporal dependencies [74]. Many of these architectural strategies (temporal recurrence, spatial convolution, and hybrid CNN–recurrent designs) were first refined on adjacent EEG-decoding tasks such as affective-state recognition, where CNN–GRU hybrids and type-2 fuzzy VAD representations have been used to model spatiotemporal EEG dynamics [75], [76]. VOLUME 11, 20235 Lightweight LSTM networks have also been developed for multi-class detection of neurodegenerative diseases from EEG, reaching up to 98% accuracy across six diagnostic categories, including Alzheimer’s, frontotemporal dementia, and vascular dementia. Being computationally efficient, they suit real-time or wearable use and widen access to deep- learning cognitive assessment tools [73]. Many of these headline accuracies, however, come from relatively small, single-site datasets; recent reviews caution that such perfor- mance often does not transfer to subject-independent, multi- site validation, underscoring the need for rigorous external testing [42]. B. CONVOLUTIONAL NEURAL NETWORKS FOR NEUROIMAGING CNNs have substantially advanced the analysis of neu- roimaging data for dementia detection [55], [56], [59]. Although some pipelines report near-ceiling accuracies ap- proaching 99.9% for Alzheimer’s disease classification from MRI [55], such values are rarely reproduced across in- dependent, peer-reviewed cohorts; more typically, CNN- based models achieve accuracies in the 97–99.7% range, de- pending on dataset quality, preprocessing, and experimental design [77]–[79]. These high figures should be interpreted with caution, as they frequently derive from small, single- site datasets with limited external validation. Advanced CNNs that add attention mechanisms (e.g., Convolutional Block Attention Modules, CBAM) and clinical variables (age, gender, cognitive scores) perform strongly [80]. The AD_Net model, based on VGG19 architecture with embed- ded Convolutional Block Attention Modules, achieved 89% accuracy [81]. C. TRANSFORMERS, GRAPH NEURAL NETWORKS, AND EEG FOUNDATION MODELS Beyond convolutional and recurrent architectures, attention- based and self-supervised models now sit at the research frontier. Vision transformers and hybrid CNN–transformer networks now rival 3D CNNs for MRI-based classification and offer attention-based saliency for feature attribution, although attention weights are an imperfect proxy for faith- ful explanation [82], [83]. Attention-based transformers that fuse information across multiple brain regions have likewise been developed for EEG decoding in adjacent affective- computing tasks, offering transferable design principles for channel- and region-aware clinical EEG classification [84]. For EEG, graph neural networks explicitly model inter- channel connectivity; frequency-band-aware graph-expert models (proposed in preprint form) have been used to recover physiologically meaningful θ/α alterations that cor- relate with clinical scores [85]. The most prominent recent development is the emergence of EEG foundation models: large networks pre-trained by self-supervision on thousands of hours of heterogeneous EEG and subsequently fine-tuned for downstream tasks. General-purpose backbones such as LaBraM and CBraMod achieve strong performance across diverse EEG benchmarks [11], [12], and an Alzheimer’s- specific foundation model pre-trained on a large multi- cohort corpus has recently been proposed in preprint form [86]. This progress nonetheless warrants caution. A public Alzheimer’s/frontotemporal-dementia EEG benchmark [87] and recent stress-testing analyses (currently preprints) [88] indicate that, under rigorous subject- and site-disjoint eval- uation, general EEG foundation models do not consistently outperform well-engineered classical features for dementia decoding, and that apparent gains can be inflated by dataset- identity confounds and subject-level data leakage. Compre- hensive reviews reach a similar conclusion, emphasizing that many reported accuracies derive from small, single- site datasets without external validation [42]. Standardized, openly benchmarked, and externally validated evaluation is therefore a prerequisite before such models can be regarded as clinically dependable. D. MULTIMODAL MACHINE LEARNING APPROACHES Integrating multiple data modalities (EEG, MRI, clinical variables, and fluid biomarkers) has become a leading way to improve the accuracy and robustness of cognitive impair- ment assessment, especially in complex neurodegenerative conditions [89], [90]. Combining EEG with structural MRI using support vector machines classifies more accurately than single-modality approaches [3]. A study comparing 29 classifiers found that the best model combined clinical data, resting-state functional MRI, and neurofilament light chain biomarkers, reaching 76.2% accuracy and 84.0% area under the curve for distinguishing cognitively normal from cognitively impaired Parkinson’s disease patients [91], [92]. More recent work extends fusion beyond simple feature con- catenation toward graph- and hypergraph-based integration and large language model–knowledge graph frameworks that combine imaging, genetic, fluid, and clinical data, in some cases across non-matched cohorts, to predict MCI-to- dementia conversion [93], [94]; a 2026 systematic review catalogs the datasets, modalities, and models now in use [43]. Early demonstrations that fuse EEG, MRI, and virtual- reality data illustrate both the promise and the added com- plexity of such comprehensive pipelines [95]. IV. DIGITAL BIOMARKERS AND EMERGING TECHNOLOGIES A. REMOTE COGNITIVE ASSESSMENT PLATFORMS The COVID-19 pandemic accelerated the development of remote cognitive assessment technologies [96], [97]. The Integrated Cognitive Assessment (ICA) is a leading example of computerized cognitive testing that can be administered remotely while retaining clinical validity. Validation studies of the ICA reported 93% sensitivity for dementia detection and 83% sensitivity for MCI detection, with 80% specificity for both [96]. The platform showed potential to reduce unnecessary referrals to memory services by correctly iden- tifying 72% of patients without cognitive impairment [96]. 6VOLUME 11, 2023 A 2025 scoping review maps the fast-growing field of remote and unsupervised digital cognitive assessments now being validated in preclinical and prodromal populations [46]. B. SPEECH, LANGUAGE, AND LARGE LANGUAGE MODEL BIOMARKERS Spontaneous speech and language have become scalable, low-burden digital biomarkers of cognitive decline: lexical, syntactic, acoustic, and semantic features degrade measur- ably in MCI and early dementia. Natural language process- ing of connected speech and clinical notes can flag decline before conventional testing [29], [32], and large language models (LLMs) have rapidly advanced this area: recent studies report accurate automated screening from spoken language [98], transformer pipelines augmented with LLM- generated synthetic data to offset limited clinical corpora [99], and systematic evaluations of transformer- and LLM- based dementia detection from speech [44]. Because such models are opaque and may encode demographic or lin- guistic bias, explainability has become a distinct research focus, with a dedicated review of explainable-AI methods for speech-based cognitive-decline detection [100]. C. VIRTUAL REALITY AND IMMERSIVE ASSESSMENT Virtual Reality (VR) offers a new approach to cognitive assessment, placing people in ecologically valid environ- ments that simulate real-world activities [101], [102]. VR- based assessments can capture subtle behavioral changes and movement patterns that traditional neuropsychological tests may miss. Pairing VR with AI enables adaptive systems that adjust difficulty in real time to a user’s performance, and these show promise for both assessment and cognitive rehabilitation [103]. Immersive VR paradigms that probe egocentric spatial memory and navigation, functions depen- dent on entorhinal–hippocampal circuits vulnerable early in Alzheimer’s disease, have distinguished amnestic MCI from healthy aging [104], and a recent systematic review and meta-analysis supports the diagnostic potential of VR-based screening while noting that most individual studies remain small and cross-sectional [45]. D. WEARABLE TECHNOLOGY AND CONTINUOUS MONITORING Wearable technologies offer new opportunities for contin- uous cognitive monitoring in everyday settings, supporting early detection and ongoing management of cognitive de- cline with data that complement traditional neuropsycholog- ical assessments. Recent systematic reviews and conceptual frameworks describe how AI-enabled digital phenotyping integrates multimodal passive-sensing streams (gait, sleep, speech, heart-rate variability, and typing dynamics) into longitudinal markers of brain health [24], [105]. These devices capture physiological signals, movement patterns, and behavioral metrics that complement standard cogni- tive assessments [106]. Consumer EEG headbands (e.g., Muse, Emotiv) and discreet behind-the-ear systems record brain activity during daily life for real-time monitoring of attention and cognitive workload; their signal-processing and home-deployment aspects are considered further in the section on portable sensing [106]. Empatica E4 is a medical- grade wrist-worn device that measures heart rate variability, electrodermal activity, skin temperature, and movement; it has been used in research studies to predict cognitive scores and track executive-function changes in people with mild cognitive impairment (MCI) [107]. Similarly, Oura Ring is a sensor-equipped ring that tracks sleep quality, heart rate, temperature, and activity patterns, providing insights into physiological changes associated with cognitive health [108]. Apple Watch and Similar Devices are also com- mercial smartwatches that can passively collect data on physical activity, sleep, heart rate, and even prompt users for cognitive tasks, supporting remote brain health assessment and early detection of MCI [109]. There are custom multisensor wearable devices integrat- ing inertial sensors (for gait and movement), microphones (for speech analysis), heart rate, and electrodermal activity sensors. Worn as wristbands or waist patches, these systems capture trends in speech, gait, and cognitive stress, support- ing early diagnosis of Alzheimer’s and related dementias [110]. Accelerometer-based devices track step count, walk- ing speed, and activity patterns, which are sensitive to early cognitive and functional decline. Examples include Fitbit and ActiGraph [111]. Table 1 summarizes representative wearable devices and the cognitive metrics they capture. V. ADVANCED SIGNAL PROCESSING, PORTABLE SENSING, AND CLINICAL VALIDATION Complementing the biomarkers and model families intro- duced above, several lines of work address the signal- processing, hardware, and validation questions that most directly govern real-world deployment; recent reviews sur- vey these signal-processing, machine-learning, and deep- learning developments for EEG-based MCI detection [112]. This section focuses on advances that are distinct from, rather than a restatement of, the preceding sections: sleep- based biomarkers, classical and signal-complexity methods, portable and home-based acquisition, and large-scale clin- ical validation. Feature-engineering strategies that jointly encode the temporal, spectral, and spatial structure of mul- tichannel EEG (for example, spatially-infused spectrogram representations) have also been proposed to make down- stream deep-learning classifiers more robust to noise [113]. A. SLEEP-BASED EEG BIOMARKERS Sleep EEG is a promising, low-burden avenue for early detection. A large-scale study using polysomnography from 8,044 participants showed that macro- and micro-structural sleep features discriminate dementia, MCI, and cognitively normal individuals, with best-model areas under the receiver operating characteristic curve of 0.78 (dementia vs. normal) and 0.73 (MCI vs. normal) [114]. Because sleep studies VOLUME 11, 20237 Device/TechnologyMain FeaturesCognitive Metrics Captured Empatica E4HRV, EDA, skin temperature, movementExecutive function, stress, arousal Oura RingSleep, HRV, temperature, activitySleep quality, circadian rhythms Apple WatchActivity, heart rate, sleep, cognitive promptsActivity/sleep patterns, cognitive tasks EEG HeadbandsBrain waves (EEG), attention, workloadNeural activity, cognitive workload Multisensor WristbandsGait, speech, HR, EDAMovement, speech, physiological stress Accelerometer TrackersSteps, walking speed, movementPhysical activity, gait variability TABLE 1. Summary of representative wearable technologies and the cognitive metrics they capture, compiled from the wearable-sensing and digital-phenotyping studies reviewed here [24], [105], [107]–[109]. are already performed routinely, such markers offer an opportunistic screening pathway. B. CLASSICAL AND SIGNAL-COMPLEXITY METHODS Alongside deep learning, classical machine learning and nonlinear signal analysis remain competitive and often more interpretable. Support vector machines applied to steady-state visual evoked potentials (EEG-SSVEP) reached 95.69% accuracy, 92.28% sensitivity, and 95.58% speci- ficity for early MCI detection in a single-site study [115], and topological data analysis with linear SVM or neural networks exceeded 90% accuracy in separating healthy aging from MCI [116]. Beyond spectral power, information- geometry classifiers for automatic ERP discrimination [117], network-synchronization measures such as phase- locking value and theta–gamma phase–amplitude coupling [118], and multifractal detrended fluctuation analysis (re- ported at 90% accuracy for normal vs. MCI) [119] cap- ture nonlinear dynamics that linear methods miss. Passive paradigms such as fast periodic visual stimulation with a lightweight CNN further reduce dependence on task effort and education, inferring impairment directly from EEG [120]. Consistent with the deep-learning results discussed earlier, these accuracies are largely single-site and require subject-independent, external validation. C. REAL-TIME, PORTABLE, AND HOME-BASED ACQUISITION Deployment outside the clinic depends on lightweight pro- cessing and comfortable hardware. Lightweight denoising networks such as LTDNet-EEG enable real-time EEG pro- cessing on resource-constrained consumer devices [121], [122]. Single-channel headbands support real-time atten- tion regulation and neurofeedback [123], portable semi- dry-electrode systems achieve signal quality comparable to medical-grade gel electrodes [124], and behind-the-ear systems improve comfort and mobility for continuous, real- world monitoring [125]. Home-based, four-channel config- urations achieve strong diagnostic performance while re- maining practical for unsupervised use, offering an objective alternative to paper-and-pencil testing [116], [119]. D. CLINICAL VALIDATION AT SCALE Larger cohorts are beginning to substantiate clinical utility. Increased P300 latency reflects disrupted large-scale net- work and working-memory function and is detectable in early, sometimes prodromal, stages [126], [127]; analysis of prefrontal ERP signals from 1,754 elderly participants revealed greater response-time and P300-latency variability in MCI, with connectivity measures serving as reliable early markers [31]. Disrupted beta-band synchronization has likewise been reported in MCI [128], and Holo-Hilbert spectral analysis achieved sensitivities and specificities of roughly 75–94% across stages of impairment [129]. Sleep- based analysis in a large polysomnography cohort further established the feasibility of opportunistic dementia screen- ing [114]. Nonetheless, most reports remain single-cohort, and prospective, multi-site validation with pre-registered endpoints is still the exception rather than the rule. VI. INTEGRATING LIFESTYLE FACTORS INTO TECHNOLOGICAL APPROACHES FOR EARLY DETECTION OF COGNITIVE DECLINE A. KEY LIFESTYLE FACTORS INFLUENCING COGNITIVE DECLINE A consistent body of evidence links modifiable lifestyle factors, including physical activity, diet quality, sleep, social engagement, vascular health, and health behaviors such as smoking and alcohol use, to cognitive trajectories in older adults [130]–[135]. Rather than re-enumerating the epidemiological strength of each factor, which is catalogued together with its protective counterpart in Section VIII and Table 4, we focus here on how these factors can be measured and acted upon technologically. Several domains yield objective, sensor-derived signals: combined physical and cognitive training reshapes functional connectivity in ways visible in EEG beta-band activity [136], and sleep can be quantified with actigraphy and polysomnography rather than by self-report alone [137]. Psychosocial and affective states that influence decline, notably depression and anxiety, can likewise be tracked continuously, since EEG-based affective-state recognition and language-model pipelines that detect depressive states from text provide scalable longitudinal monitoring of these risk factors [138]– [140]. B. THE ROLE OF AI AND ML IN INTEGRATING LIFESTYLE DATA AI and ML models can process large, heterogeneous datasets (combining lifestyle, clinical, genetic, and neuro- physiological variables) to build predictive models of cogni- tive decline. Such models can identify high-risk individuals 8VOLUME 11, 2023 from baseline lifestyle factors such as smoking, physical activity, sleep, diet, and social participation [134]; support personalized prevention by tailoring dietary or physical- activity recommendations to individual risk profiles [131]; integrate wearable EEG and lifestyle-monitoring streams to provide continuous, real-time assessment and feedback that enhance early-intervention opportunities [136], [141]; and combine clinical, lifestyle, and neurophysiological variables into risk models that achieve high predictive accuracy in specific populations, such as older adults with diabetes [130]. C. TECHNOLOGICAL AND CLINICAL IMPLICATIONS These capabilities translate into several practical implica- tions. As digital biomarkers, EEG and wearable devices can objectively quantify the impact of lifestyle interventions (for example, exercise or cognitive training) on brain function, providing measurable endpoints for prevention programs [136]. For remote monitoring, AI-powered platforms can integrate self-reported lifestyle, sleep tracking, and dietary logs with neurophysiological data to enable scalable, home- based cognitive-health monitoring [137], [141]. In the ser- vice of precision medicine, ML algorithms can tailor pre- vention and intervention strategies to an individual’s partic- ular combination of lifestyle factors, genetic profile (e.g., APOE ε4 status), and brain biomarkers [131], [142]. D. SUMMARY Early detection and prevention of cognitive decline in older adults should combine technological advances (EEG, AI, ML) with modifiable lifestyle factors. Together these sup- port more accurate risk prediction, personalized interven- tion, and, ultimately, better cognitive-health outcomes for aging populations. VII. AN INTEGRATIVE FRAMEWORK FOR EARLY DETECTION OF COGNITIVE IMPAIRMENT The preceding sections surveyed individual technologies; this section synthesizes them from the standpoint of early detection, where the practical question is not which single modality is most accurate but how complementary signals can be combined into scalable, clinically actionable screen- ing pathways (Fig. 3). Early identification has acquired new urgency because the therapeutic and preventive window is widest before overt dementia: disease-modifying anti- amyloid therapies are indicated in early-stage disease [6], [8], and a substantial fraction of dementia (estimated at up to 45%) is potentially preventable through modifiable risk factors that are most tractable earlier in life [10]. A. THE PRECLINICAL AND PRODROMAL WINDOW Alzheimer’s disease and related dementias evolve along a continuum, from a preclinical, biomarker-positive but asymptomatic phase, through mild cognitive impairment (MCI), to dementia. The 2024 revised diagnostic framework formalizes this biological staging and admits fluid biomark- ers as evidence of underlying pathology [9], shifting the goal of technology from confirming established dementia toward detecting subtle deviations years before functional decline. This biological reframing is not universally endorsed: the International Working Group cautions against diagnosing Alzheimer’s disease in cognitively unimpaired individuals on the basis of biomarkers alone, given the imperfect and age-dependent association between biomarker positivity and subsequent clinical decline [143]. In older adults, moreover, cognitive impairment frequently reflects mixed pathology (co-occurring Alzheimer’s, cerebrovascular, Lewy-body, and TDP-43 changes), so a single proteinopathy marker rarely captures the full substrate [144], and differentiating demen- tia subtypes, including potentially reversible contributors such as depression, medication effects, and metabolic dis- turbance, remains an essential clinical task that technology should support rather than supplant. Because only a minority of MCI cases progress in any given year [145], effective early detection must not merely classify current status but stratify future risk, which in turn demands longitudinal and multimodal evidence rather than a single cross-sectional test. B. A TIERED, MULTIMODAL SCREENING STRATEGY No single modality is simultaneously inexpensive, scalable, and definitive, which motivates tiered strategies (Fig. 4) in which accessible first-tier tools triage who should re- ceive costlier confirmatory assessment. Low-burden, high- throughput instruments (computerized and remote cognitive tests [46], [96], speech and language analysis [32], [98], and, increasingly, plasma biomarkers usable in primary care [146]) can flag individuals who warrant second-tier characterization with EEG, structural or molecular imaging, or specialist evaluation. Opportunistic screening is particu- larly attractive: routinely collected polysomnography can be repurposed for dementia risk assessment [114], and a single primary-care blood draw can now provide amyloid-status in- formation that previously required positron-emission tomog- raphy or lumbar puncture [13], [147]. Table 3 compares the principal detection methods across accuracy, invasiveness, and clinical readiness. C. COMPLEMENTARITY OF MODALITIES FOR RISK STRATIFICATION The modalities surveyed here capture distinct facets of the same underlying process and are therefore complemen- tary rather than competing. Molecular assays (plasma p- tau217, Aβ42/40) index proteinopathy, with cerebrospinal- fluid Aβ42/40 and p-tau, and amyloid/tau PET (the latter enabling Braak-like staging), serving as established confir- matory standards [9], [148]; structural and diffusion MRI quantify neurodegeneration and network disconnection [3]; EEG and event-related potentials expose functional and connectivity disturbances that can precede structural loss [31]; and digital and behavioral markers (gait, speech, sleep, typing dynamics) reflect the real-world consequences of VOLUME 11, 20239 Acquisition (EEG, MRI, blood, speech, wearables) Preprocessing & artifact removal Feature / representation learning Model (ML, CNN/RNN, transformer, fusion) Validation (subject- & site-independent) Clinical decision & intervention longitudinal monitoring and re-assessment FIGURE 3. A representative detection-to-management pipeline for cognitive impairment. Heterogeneous signals are acquired, preprocessed, and transformed into features or learned representations; models are trained and, critically, validated under subject- and site-independent protocols before informing clinical decisions and interventions, with longitudinal monitoring closing the loop. Tier 1: Population-scale screening digital cognitive tests, speech/language, wearables Tier 2: Risk stratification digital markers+ risk factors (age, APOE, lifestyle) Tier 3: Blood biomarkers (primary care) p-tau217, Aβ42/40 ratio Tier 4: Confirmatory imaging/CSF MRI, amyloid/tau PET (specialist) Tier 5: Diagnosis & matched intervention Longitudinal monitoring FIGURE 4. A tiered, multimodal framework for early detection. Low-cost, population-scale tools (Tiers 1–2) triage who proceeds to blood-based biomarkers in primary care (Tier 3) and, if warranted, to confirmatory imaging or CSF at specialist centers (Tier 4), coupling diagnosis to matched intervention (Tier 5); longitudinal monitoring closes the loop. these changes [105]. Consequently, multimodal fusion often improves early sensitivity and specificity relative to any single stream [91], [93], although it can also add complexity and noise, and systematic evidence now catalogs the datasets and model families that make such fusion reproducible [43]. Combining functional (EEG) and structural (MRI) information, for instance, has been shown to separate stages of vascular cognitive impairment more accurately than either modality alone [3]. D. DIGITAL-FIRST AND REMOTE EARLY SCREENING AT SCALE Population-level early detection requires instruments that operate outside specialist clinics. Smartphone- and wearable-based digital phenotyping enables continuous, ecologically valid monitoring of cognition-relevant behav- ior [24]; speech-based and large-language-model pipelines allow unsupervised screening from brief spoken samples or routine clinical notes [29], [98]; and immersive virtual- reality tasks probe spatial-navigation deficits that are sen- sitive to early entorhinal–hippocampal involvement [104]. Such tools can extend screening to underserved and remote populations and can accelerate clinical-trial recruitment [92], [149]; their principal limitations, validation against biomarker-confirmed cohorts and equitable performance across languages and cultures, are considered below. E. COUPLING EARLY DETECTION TO TIMELY INTERVENTION Detection is clinically meaningful only when it is linked to action. A positive early-detection result can trigger modifiable-risk-factor management and structured multido- main lifestyle programs, which have demonstrated cognitive benefit in large randomized trials [14], [150], [151]; identify candidates for, and support the safety monitoring of, disease- modifying therapy in appropriately staged patients [8]; and establish a personalized longitudinal baseline against which subsequent change can be measured [152]. This ‘detect- then-act’ loop, rather than one-off classification, is where the integration of the technologies surveyed here is most likely to yield public-health benefit. F. BARRIERS SPECIFIC TO EARLY DETECTION Screening asymptomatic or minimally symptomatic popula- tions raises challenges distinct from diagnosing established disease. Where progression base rates are low, even accurate tests yield limited positive predictive value, so confirmatory testing and longitudinal follow-up are essential to avoid over-diagnosis. Heterogeneous protocols and the absence of standardized, externally validated pipelines continue to im- pede comparison and deployment [153], and the disclosure of pre-symptomatic risk carries ethical and psychological implications that must be managed through appropriate consent and support [33]. Addressing these issues (through standardization, prospective multi-site validation, and equi- table, privacy-preserving implementation) is a prerequisite for translating early-detection technologies into routine care. VIII. RISK AND PROTECTIVE FACTORS FOR MCI TO DEMENTIA PROGRESSION Approximately 3–15% of individuals with mild cognitive impairment (MCI) progress to dementia each year [145]; cumulative progression is substantially higher over long- term follow-up [154]. Understanding the risk and protective factors that shape this progression is central to designing early interventions that can delay or prevent dementia onset. Table 4 summarizes the principal risk and protective factors together with the strength of their supporting evidence. A. KEY RISK FACTORS WITH STRONGEST EVIDENCE Genetic and Biomarker Risk Factors APOE ε4 allele is the strongest genetic risk factor, with a pooled odds ratio of approximately 2.3 for progression [66], [160]. Notably, sex differences exist, with APOE ε4 status and depression appearing to increase progression risk more significantly in females. Conversely, APOE ε2 status provides protective effects, particularly in males [66], [162]. 10VOLUME 11, 2023 Intervention TypeTarget PopulationMechanism of ActionEvidence LevelEffectivenessImplementation ChallengesTime to Benefit Physical ExerciseAt-risk adults, MCI patientsNeuroplasticity, BDNF, vascular health Strong∼14–21% lower risk [155] Adherence, individual variability 3–6 months Cognitive TrainingMCI patients, early dementia Cognitive reserve, synaptic plasticity ModerateImproves trained domains; transfer uncertain Engagement, transfer to daily life 6–12 months Mediterranean DietGeneral population, at-risk adults Anti-inflammatory, antioxidant effects Strong∼27–36% lower risk [156] Cultural adaptation, sustainability 1–2 years Social EngagementOlder adults, isolated individuals Cognitive stimulation, stress reduction ModerateProtective association [10] Social infrastructure, accessibility 6–12 months Blood Pressure ControlHypertensive patientsVascular protection, reduced tau Strong∼15–19% lower risk (MCI) [157] Medication adherence, side effects 1–2 years Diabetes ManagementDiabetic patientsMetabolic optimization, inflammation StrongRisk factor; glycaemic benefit unproven [10] Lifestyle changes, monitoring 6–12 months Hearing Aid UseHearing-impaired older adults Cognitive load reduction, social engagement Moderate∼19% lower risk [158] Stigma, cost, access3–6 months Sleep OptimizationSleep disorder patientsAmyloid clearance, memory consolidation ModerateEmerging; RCT evidence limited Diagnosis, treatment compliance 3–6 months Smoking CessationSmokersVascular health, inflammation reduction StrongRisk declines after cessation [10] Addiction, support systemsImmediate + long-term Digital TherapeuticsMCI patients, at-risk adultsCognitive training, behavior modification Emerging∼30% (simulation estimate) [159] Technology literacy, validation 3–6 months Pharmacological (Aducanumab) Early AD patients, amyloid-positive Amyloid plaque reductionLimitedWithdrawn from market 2024 [7] Cost, limited efficacy, availability 12–18 months Pharmacological (Lecanemab) Early AD patients, amyloid-positive Amyloid plaque reductionHigh (RCT); benefit contested ∼27% slowing [6]Cost, availability, selection12–18 months Combined Lifestyle Programs At-risk populationsSynergistic effects across domains StrongModest RCT cognitive benefit [14] Coordination, long-term commitment 1–3 years TABLE 2. Comparison of interventions for cognitive impairment prevention and management. Reported effect sizes are indicative and drawn from the cited, heterogeneous primary sources; qualitative entries denote a recognized association without a single well-established quantitative estimate. The 45% figure is a population-attributable fraction for all modifiable factors combined (a population ceiling), not a demonstrated effect of lifestyle programs [10]. Detection MethodSpecific MeasuresAdvantagesLimitationsClinical Readiness Plasma biomarkersAβ42/40 ratio, p-tau217, p-tau181, GFAP, NfL Non-invasive, cost-effective, accessible Newer technology, standardization needed Emerging (2024–2025) [13], [148] CSF biomarkersAβ42, total tau, p-tau181, p-tau217 High accuracy, established protocols Invasive procedure, specialized centers Established Neuroimaging (MRI)Hippocampal volume, cortical thickness Structural brain changes, widely available Expensive, requires expertiseEstablished Neuroimaging (PET)Amyloid PET, tau PET, FDG PET Direct pathology visualization Very expensive, limited availability Established Cognitive assessmentMMSE, MoCA, memory tests Easy to administer, standardized Subjective, practice effectsEstablished Functional assessmentFAQ, ADL assessmentsReal-world impact assessment Depends on informant reliability Established Genetic testingAPOE genotypingRisk stratificationLimited predictive value alone Established Digital biomarkersSmartphone-based cognitive tests Continuous monitoring, convenient Validation needed, technology barriers Developing Combined assessmentMulti-modal assessment panels Comprehensive risk assessment Complex, time-consuming, expensive Developing TABLE 3. Comparison of cognitive-impairment detection methods, compiled from the biomarker, neuroimaging, and clinical-assessment literature reviewed here [9], [13], [67], [147], [148]. Blood-based biomarkers have advanced rapidly from an emerging tool to a clinically actionable one. Phosphorylated tau-217 (p-tau217) demonstrates the strongest predictive capability, with area under the curve (AUC) values of 0.926–0.946 [161], and independent studies report accuracy comparable to cerebrospinal fluid and approaching amyloid PET, outperforming routine primary-care clinical assessment [146]–[148]. The plasma Aβ42/40 ratio can reach AUCs near 0.94 against amyloid PET on high-sensitivity platforms, although its accuracy is strongly assay-dependent [161], [163], and p-tau181 offers good but generally lower accu- racy than p-tau217 [163]. Table 5 summarizes representative blood-biomarker performance and clinical readiness. Re- flecting this maturation, the 2024 revised Alzheimer’s Asso- ciation criteria formally incorporate blood-based biomarkers into diagnosis and staging [9], and in May 2025 the U.S. Food and Drug Administration cleared the first blood test (a plasma p-tau217/Aβ42 ratio assay) to aid Alzheimer’s diag- nosis in symptomatic adults [13]. Fully automated platforms and community-cohort validation are extending these assays toward routine and prognostic use [164], [165], although standardized cut-points and cross-population validation re- main works in progress [166]. Neuroimaging Predictors Hippocampal volume reduc- tion consistently predicts progression from MCI to dementia [66]. Additional structural changes include reduced gray matter volume in bilateral cerebellar cortices, temporal cortex, and insular cortices. These neuroimaging markers are particularly valuable as they provide objective measures of brain pathology [67]. VOLUME 11, 202311 Factor CategoryFactorTypeStrength of Evidence ModifiableEffect Size/Comments Cognitive/Neuropsychological Cognitive reserveProtectiveStrongYesHigher reserve delays progression Memory impairment severityRiskStrongNoSeverity correlates with progression MMSE scoreRiskStrongNoLower scores predict progression Global Deterioration Scale (GDS)RiskStrongNoHigher scores predict progression Functional Activities Questionnaire (FAQ) RiskStrongNoHigher scores predict progression DemographicOccupational attainmentProtectiveStrongNoHigher levels protective AgeRiskStrongNoAdvancing age increases risk SexRiskStrongNoFemale sex shows different patterns Genetic/BiomarkersAPOE ε2 alleleProtectiveModerateNoProtective for males APOE ε4 alleleRiskStrongNoOR≈2.3 [160] Clusterin (CLU) geneRiskModerateNoIndependent risk factor Amyloid-β42 (Aβ42)RiskStrongNoLower levels predict progression Phosphorylated tau (p-tau181)RiskStrongNoHigher levels predict progression Phosphorylated tau (p-tau217)RiskStrongNoStrongest single biomarker; AUC 0.926–0.946 [161] Tau protein (total)RiskStrongNoElevated levels predict progression Neurofilament light chain (NfL)RiskModerateNoAssociated with neurodegeneration Lifestyle/BehavioralPhysical inactivityRiskStrongYesMost modifiable risk factor SmokingRiskStrongYesIncreases risk significantly Poor dietRiskStrongYesMediterranean diet protective Excessive alcoholRiskModerateYesModerate consumption may be protective Sleep disordersRiskModerateYesSleep apnea increases risk ObesityRiskStrongYesMidlife obesity increases risk Medical/VascularCardiovascular diseaseRiskStrongPartiallyMultiple vascular factors increase risk DiabetesRiskStrongYesSignificantly increases risk HypertensionRiskStrongYesMidlife hypertension particularly risky Stroke historyRiskStrongPartiallyIncreases risk, especially in males Hearing lossRiskStrongYesMidlife hearing loss increases risk NeuroimagingHippocampal volumeRiskStrongNoReduced volume predicts progression MRI markersRiskStrongNoStructural changes predict progression Gray matter volumeRiskModerateNoReduced volume in key regions PsychosocialMarital statusProtectiveModerateNoMarriage protective for females Educational levelProtectiveStrongNoHigher education protective DepressionRiskStrongYesIncreases risk, especially in females Social isolationRiskStrongYesIncreases risk significantly TABLE 4. Summary of risk and protective factors for progression from MCI to dementia, compiled from the sources reviewed here [10], [66], [160], [161]. TABLE 5. Representative diagnostic performance of blood-based biomarkers for Alzheimer’s disease. AUC values are as reported in the cited primary studies; the reference standard differs across markers (amyloid/tau PET versus clinical diagnosis), and the plasma Aβ42/40 ratio in particular is strongly assay-dependent. Blood biomarkerDiscriminatesAUCClinical readiness / notes p-tau217Amyloid-/tau-PET status0.92–0.96Strongest single plasma marker; component of the FDA-cleared test [147], [148] p-tau217/Aβ42 ratioAD pathology (amyloid)0.95–0.97FDA-cleared plasma test (2025); high accuracy in primary care [13], [146], [161] Aβ42/40 ratioAmyloid-PET status∼0.94Included in the cleared ratio test; standalone accuracy is assay-dependent [161], [163] p-tau181AD vs. controls∼0.88Earlier-generation p-tau; lower accuracy than p-tau217 [163] GFAPAmyloid-positive vs. negative0.69–0.86Astrocytic-reactivity marker; sensitive but not AD-specific [167] NfLAD dementia vs. controls∼0.87Nonspecific marker of neurodegeneration [168], [169] Cognitive and Functional Assessments Memory impair- ment severity, particularly performance on the Rey Auditory Verbal Learning Test (RAVLT) delayed recall, strongly pre- dicts progression. The Functional Activities Questionnaire (FAQ) also serves as a robust predictor, with higher scores indicating greater risk [66], [170]. B. MOST SIGNIFICANT PROTECTIVE FACTORS Cognitive Reserve Higher educational attainment and oc- cupational complexity provide substantial protection against cognitive decline [162], [171]. These factors contribute to cognitive reserve, which allows individuals to maintain function despite underlying brain pathology. Lifestyle Factors Physical exercise emerges as the most modifiable protective factor, with meta-analytic estimates of 12VOLUME 11, 2023 roughly 15–20% lower dementia risk [155], [172], [173]. The protective effects occur through multiple mechanisms including neuroplasticity enhancement, BDNF upregulation, and vascular health improvement [173]. Adherence to Mediterranean and MIND dietary patterns is associated with an estimated 27–36% lower risk of cognitive impairment and Alzheimer’s disease [156], [174], while social engagement is a recognized protective factor and social isolation a modifiable risk factor [10], [175]. Be- ing married also appears particularly protective for women [162]. C. EVIDENCE-BASED EARLY INTERVENTION STRATEGIES Multimodal Lifestyle Interventions Multidomain lifestyle programs modeled on the FINGER trial can improve or stabilize cognition across several domains when multiple risk factors are addressed simultaneously [150], [151]. The FINGER trial model has been adapted globally, showing that interventions targeting diet, exercise, cognitive training, and vascular health can significantly reduce dementia risk [151]. Most recently, the large U.S. POINTER randomized trial reported that a structured multidomain lifestyle intervention improved global cognition relative to a self-guided program in at-risk older adults, providing the strongest contemporary evidence that behavioral prevention is achievable at scale [14]. Table 2 compares the major intervention types, their mechanisms, evidence levels, and reported effect sizes. Cardiovascular Risk Management In the SPRINT MIND randomised trial, intensive blood-pressure control re- duced the risk of MCI by approximately 15–19% (although the primary probable-dementia endpoint did not reach sig- nificance) [157]; diabetes is an established modifiable risk factor, although randomised trials of intensive glycaemic control have not shown a clear cognitive benefit [10]. These interventions are thought to act through vascular protection and reduced tau pathology. Physical Exercise Programs Aerobic exercise combined with resistance training shows optimal benefits. The protec- tive effects are dose-dependent, with moderate to vigorous intensity activities providing greater cognitive benefits than low-intensity exercise [176]. Precision Medicine Approaches Personalized interven- tions based on individual risk profiles show promise. Risk stratification that incorporates genetic information, biomark- ers, and clinical assessments enables targeted interventions. Recent studies suggest that interventions should be tailored to individual performance levels and risk factors [150], [177]. Neuromodulation and Cognitive Rehabilitation Be- yond pharmacology and lifestyle, technology-enabled neu- romodulation is being evaluated as a management option. Home-based transcranial direct-current stimulation has pro- duced short-term cognitive gains in MCI [27], and transcra- nial magnetic stimulation combined with EEG is used both to probe and to modulate cortical excitability across the Alzheimer’s continuum [28]. Sensory and 40-Hz (gamma) transcranial alternating-current stimulation seeks to entrain the oscillatory activity disrupted early in disease, with randomized, sham-controlled trials now reporting effects on cognition and neural markers [65]. In parallel, immer- sive virtual-reality and related neurotechnological platforms are being adapted for cognitive rehabilitation as well as assessment [102], and prescription digital therapeutics are increasingly positioned as adjuncts that may complement pharmacological treatment [178]. The evidence base remains early-stage, however, and durable, clinically meaningful benefit will require confirmation in larger, longer, and adequately controlled trials. D. IMPLEMENTATION CHALLENGES AND SOLUTIONS Digital Health Solutions Prescription digital therapeutics (PDTs) targeting multiple risk factors simultaneously show 30% relative risk reduction in simulation studies [159]. Mobile health interventions with remote coaching can reach underserved populations, though implementation challenges include technology literacy and sustained engagement [179]. Healthcare System Integration Early screening pro- grams integrated into routine medical check-ups are es- sential [180]. The Kobe Project shows how community- based implementation can pair early detection with broader intervention programs [181]. E. SUMMARY AND RECOMMENDATIONS The progression from MCI to dementia is influenced by multiple interconnected factors, with the strongest risk pre- dictors being genetic markers (APOE ε4), biomarkers (p- tau217, Aβ42/40 ratio), and structural brain changes. Early detection using emerging blood-based biomarkers combined with multimodal lifestyle interventions offers the most promising approach for preventing or delaying dementia onset. Primary Prevention Focus Evidence supports shifting focus from treatment to prevention: the 2024 Lancet Com- mission identifies fourteen modifiable risk factors, adding untreated vision loss and elevated LDL cholesterol to its earlier list, and estimates that nearly 45% of dementia cases are potentially preventable by addressing them across the life course [10]. Primary prevention strategies yield larger long-term benefits than secondary interventions targeting established pathology [172], [182]. Emerging Therapeutic Approaches Amyloid-targeting monoclonal antibodies have reached clinical practice: lecanemab (pivotal CLARITY-AD trial; U.S. Food and Drug Administration traditional approval in 2023) and donanemab (TRAILBLAZER-ALZ 2; approval in 2024) both slow decline in early-stage disease, with relative slowing on the order of 27–35% (reported on different primary instruments, CDR-SB for lecanemab and the iADRS for donanemab, so the two percentages are not directly comparable) but small absolute differences on clinical rating scales and a material risk of amyloid-related imaging abnormalities (ARIA), whose incidence is markedly higher in APOE ε4 VOLUME 11, 202313 homozygotes and which mandates genotype-informed coun- seling and serial MRI surveillance [6], [8]. Long-term extension data are accruing [183], [184], yet several authors caution that the evidence for genuine disease-course mod- ification remains contested and that effect sizes may fall below thresholds of clinical meaningfulness [185], [186]. In parallel, digital therapeutics combined with traditional interventions offer potentially synergistic benefits [178]. Population-Level Interventions Public health ap- proaches addressing modifiable risk factors at the population level show promise for reducing overall dementia burden [187]. These include policies promoting physical activity, healthy diets, and social engagement across the lifespan. IX. CLINICAL TRANSLATION AND IMPLEMENTATION CHALLENGES A. DATASETS, BENCHMARKS, AND EVALUATION Progress in this field is shaped by the datasets and evaluation protocols on which methods are trained and compared. For neuroimaging, large public cohorts such as ADNI, OASIS, and NACC provide harmonized MRI, PET, fluid-biomarker, and clinical data; for EEG, resources including the Temple University Hospital EEG Corpus, the Korean CAUEEG de- mentia dataset, and the openly released AHEPA Alzheimer’s and frontotemporal-dementia recordings are increasingly used, and a dedicated benchmark has been proposed to stan- dardize machine-learning evaluation on the latter [87]; for speech, the DementiaBank and ADReSS/ADReSSo chal- lenge corpora underpin much of the language-based work. A recent systematic review catalogs the datasets, modalities, and models now in common use for multimodal Alzheimer’s diagnosis [43]. Reported performance, however, depends critically on evaluation design. Comprehensive reviews and stress-tests emphasize that many headline accuracies arise from small, single-site datasets evaluated with subject-dependent cross- validation, in which epochs from the same individual appear in both training and test folds (illustrated in Fig. 5); under rigorous subject- and site-disjoint validation, performance typically drops substantially, and models can inadvertently learn dataset identity rather than pathology [42], [88]. Subject-independent training and testing, in which no data from a test individual appear during training, has been demonstrated for related EEG-decoding tasks and provides a practical template for such evaluation [188]. Sound re- porting therefore requires subject-independent and, ideally, external multi-site validation; attention to class imbalance and to metrics beyond raw accuracy (balanced accuracy, area under the ROC curve, sensitivity, and specificity); and transparent handling of data leakage. Standardized, openly benchmarked evaluation is a prerequisite for meaningful cross-study comparison and for eventual regulatory accep- tance. Table 6 juxtaposes representative reported figures with their evaluation designs, illustrating that the highest headline accuracies tend to originate from small, single- site studies without external validation, whereas the largest cohorts and independent stress-tests report more measured performance. B. STANDARDIZATION AND VALIDATION ISSUES Despite promising research results, several challenges im- pede the clinical translation of AI-based cognitive as- sessment tools [189]. Standardization of EEG protocols, feature-extraction methods, and validation procedures re- mains a major barrier to adoption [153]. The heterogeneity of study populations, recording conditions, and analysis approaches makes it hard to compare results across research groups. This variability necessitates larger, more standard- ized validation studies to establish clinical utility. Beyond larger cohorts, adherence to AI-specific reporting standards (such as TRIPOD+AI for clinical prediction models [190] and CLAIM for medical-imaging AI [191]), together with routine release of code, trained weights, and data-access conditions, would materially improve transparency and re- producibility. Evaluation should likewise report probability calibration (for example, calibration curves and expected calibration error) and predictive uncertainty (for example, conformal prediction or deep ensembles), because a well- calibrated, uncertainty-aware model is essential for safe de- cisions at the low base rates typical of population screening. Achieving the cross-site robustness that external validation demands further motivates domain-adaptation and domain- generalization techniques, such as covariance/Riemannian alignment and adversarial feature alignment for cross- subject and cross-site EEG transfer [192]; for regulated deployment, AI as a software medical device additionally requires a predetermined change-control plan that distin- guishes locked from continually learning models. C. EXPLAINABILITY AND CLINICAL ACCEPTANCE The "black box" nature of many AI models poses challenges for clinical acceptance [59]. Healthcare providers require interpretable results to make informed clinical decisions, driving the development of explainable AI frameworks for cognitive assessment. Recent efforts have focused on devel- oping transparent CNN models that can identify specific brain regions associated with cognitive impairment [90]. Complementary feature-selection and post-hoc explainabil- ity approaches (for example, genetic-algorithm-based chan- nel and feature optimization combined with explainable-AI attribution for EEG classification) further improve trans- parency without sacrificing accuracy [193]. D. DATA PRIVACY AND ETHICAL CONSIDERATIONS The use of AI in healthcare raises important privacy and ethical concerns, particularly for vulnerable elderly pop- ulations. Ensuring data security, informed consent, and equitable access to these technologies requires careful con- sideration of regulatory and ethical frameworks. Privacy- preserving paradigms such as federated learning, which trains shared models across institutions without exchanging raw patient data, are being adapted to dementia applications, 14VOLUME 11, 2023 TABLE 6. Representative reported performance for technology-based cognitive-impairment detection, alongside the evaluation design that conditions how each figure should be interpreted. Figures are as reported in the cited primary sources; the highest single-dataset accuracies have generally not been subjected to subject- and site-independent external validation and should be read accordingly. StudyModality / methodCohort / datasetEvaluation designReported performance MRI 3D-CNN [55]Structural MRI, deep CNNSingle 4-class datasetSingle dataset; external validation not reported 99.94% accuracy (4-class) EEG–LSTM [2]EEG, LSTMSingle-siteExternal validation not reported96.41% acc; 96.55% sens; 95.95% spec EEG-SSVEP–SVM [115]EEG-SSVEP, SVMSingle-siteExternal validation not reported95.69% acc; 92.28% sens; 95.58% spec Topological data analysis [116] EEG, TDA + SVM/NNSingle-siteExternal validation not reported>90% accuracy (HC vs. MCI) Multifractal DFA [119]EEG, MFDFASingle-siteExternal validation not reported90% accuracy (HC vs. MCI) Holo-Hilbert analysis [129] EEG, Holo-Hilbert spectrumSingle cohortCross-validation (split not specified) 75–94% sens/spec across stages Sleep-EEG screening [114] Polysomnography features8,044 participantsLarge single cohortAUC 0.78 (dementia); 0.73 (MCI) Multimodal fusion [91]Clinical + rs-fMRI + NfLPD cohort; 29 classifiersCross-validation (split not specified) 76.2% acc; 84.0% AUC ERP–connectivity [31]EEG/ERP (P300)1,754 participantsLarge cohort; group-levelSignificant RT / P300-latency markers VR-based screening [45]Virtual realityMeta-analysis, 29 studiesPooled across studiesSens 0.883; spec 0.887 EEG foundation models [11], [86] Self-supervised pre-trainingMulti-cohortSubject/site-disjoint stress-tests [87], [88] No consistent gain over classical features (a) Subject-dependent CV Train: epochs of S1–S8 Test: other epochs of S1–S8 Same subjects in both folds ⇒ leakage, inflated accuracy (b) Subject-independent CV Train: S1–S6 (Site 1) Test: S7–S8 (Site 1) Disjoint subjects, one site ⇒ fairer, still single-site (c) External / cross-site Train: Site A cohort Test: Site B cohort Different site/device ⇒ most rigorous, realistic Increasing validation rigour −→ lower but more trustworthy reported performance FIGURE 5. Evaluation designs and their effect on reported performance. Under subject-dependent cross-validation (a), epochs from the same individuals populate both training and test folds, allowing models to exploit subject identity and inflating accuracy; subject-independent (b) and, most stringently, external cross-site validation (c) progressively remove this leakage and yield lower but more trustworthy estimates. for example a federated vision-transformer framework with built-in explainability for Alzheimer’s staging [194]. Real- istic federated deployment must nonetheless contend with non-independent and heterogeneously distributed (non-IID) data across sites, communication constraints, and residual privacy leakage, motivating differential privacy and secure aggregation. Equitable deployment additionally requires ex- plicit measurement of algorithmic bias (subgroup-stratified sensitivity, specificity, and calibration across age, sex, an- cestry, and language) rather than aggregate accuracy alone, a concern especially acute for speech- and language-based markers [100]. X. FUTURE DIRECTIONS AND RESEARCH OPPORTUNITIES Several research priorities recur across the modalities sur- veyed above. Continued multimodal integration (combin- ing EEG, neuroimaging, digital biomarkers, and behavioral assessment) remains foundational [68], [89], [91], and we focus here on the further directions it enables. A. PERSONALIZED MEDICINE APPROACHES Personalized cognitive assessment and intervention is a key future direction [195]. AI systems that can adapt to individual patient characteristics, genetic profiles, and risk factors may enable more precise and effective interventions. Computational-neuroscience models of synaptic plasticity and systems-biology simulations further offer mechanistic tools that could inform drug discovery and individualized treatment planning in neurodegenerative disorders [196], [197]. B. LONGITUDINAL MONITORING AND PREDICTION Advances in continuous monitoring technologies and pre- dictive modeling offer opportunities for early intervention before clinical symptoms become apparent [152]. Longitu- dinal studies incorporating multiple assessment modalities will be essential to build these predictive capabilities. C. POPULATION-SPECIFIC CONSIDERATIONS Future research must address the need for population- specific validation of AI-based assessment tools [56]. Cul- tural, linguistic, and socioeconomic factors may influence cognitive assessment results, requiring diverse and rep- resentative validation studies. Cross-cultural analyses of emotional EEG functional networks, for instance, reveal measurable differences in connectivity between Eastern and Western populations, underscoring the need to validate VOLUME 11, 202315 neurophysiological markers across cultural groups [198]. D. EMERGING COMPUTATIONAL PARADIGMS Quantum-enhanced machine learning has been proposed for computationally intensive EEG analysis, offering theoretical speed-ups for specific subproblems. Empirical evidence of benefit for cognitive-impairment detection is, however, currently lacking, and its near-term clinical impact remains speculative [125], [199]. XI. LIMITATIONS OF THIS REVIEW Several limitations should be borne in mind when inter- preting this survey. First, as a narrative (critical) review rather than a systematic or scoping review, it does not report an exhaustive, protocol-driven search with screening counts; although major cross-disciplinary databases were queried, relevant studies may have been missed, and in- clusion reflects the authors’ judgement of representative- ness. Second, the field is advancing rapidly, and a small number of the most recent works cited, particularly certain EEG foundation-model studies, are preprints that have not completed peer review; their findings should be regarded as provisional and are labeled accordingly. Third, many reported diagnostic accuracies originate from small, single- site datasets with heterogeneous preprocessing and limited external validation, which hampers direct comparison and likely inflates apparent performance; we have therefore highlighted, rather than pooled, such figures. Fourth, the quantitative values summarized in the comparison tables are drawn from heterogeneous primary sources and are intended to be indicative rather than definitive. Finally, much of the evidence originates from high-income settings and predominantly one ancestry group, limiting generalizability; cross-population, multilingual, and longitudinal validation therefore remains a priority. XII. CONCLUSION This review examined how neurophysiological, neuroimag- ing, molecular, and digital technologies, coupled with ar- tificial intelligence, are reshaping the early detection and management of cognitive impairment in older adults. Rather than treating each modality in isolation, we organized the evidence into a single cross-disciplinary taxonomy, ap- plied a consistent methodological-rigor lens that foregrounds subject- and site-independent validation, proposed an inte- grative early-detection framework that links tiered screening to timely intervention, and consolidated the literature into comparison tables of detection methods, interventions, and risk and protective factors. Several conclusions emerge. Electroencephalography and neuroimaging, analyzed with deep and increasingly self- supervised foundation models, can separate healthy aging, MCI, and dementia with high reported accuracy [2], [114], although these figures are often obtained on small, single- site datasets and tend to fall under subject- and site- independent evaluation [42], [88]. Molecular biomarkers have advanced fastest: plasma phosphorylated tau-217 now approaches the accuracy of cerebrospinal-fluid and imaging references [146], [148], and the first blood test to aid Alzheimer’s diagnosis was cleared in 2025 [13]. Disease- modifying anti-amyloid therapies have entered the clinic with modest and still-debated benefits [6], [8], whereas multidomain lifestyle programs have supplied the strongest recent evidence that a substantial and potentially modifiable share of dementia risk can be addressed [10], [14]. At the same time, wearable, remote, speech and language, and virtual-reality tools are moving assessment beyond the clinic toward continuous, ecologically valid monitoring [24], [46], and the fusion of complementary modalities generally improves sensitivity and specificity over any single stream [43], [91]. The route from these advances to routine care remains demanding. Progress is held back by the absence of stan- dardized acquisition and analysis protocols, which makes results hard to compare across studies; by the limited interpretability of complex models, which slows clinical acceptance and motivates explainable AI [59], [83]; and by unresolved questions of data privacy, equity, and access that determine whether these tools reach the populations most at risk. A single methodological message recurs throughout the literature: headline accuracies must be tempered by subject- independent and, ideally, external multi-site validation, with attention to class imbalance and to metrics beyond raw accuracy [42], [188]. The near-term future of cognitive healthcare therefore lies less in any single breakthrough than in integration. The most promising path fuses EEG, neuroimaging, blood-based and digital biomarkers, and modifiable lifestyle factors into trustworthy, longitudinally validated predictive systems that couple early detection to personalized and actionable intervention [68], [89]. Realizing this vision will require privacy-preserving and federated learning, transparent and clinically interpretable models, and cross-population, multi- lingual, and longitudinal validation. If these conditions are met, the technologies surveyed here could shift cognitive care from late, reactive diagnosis toward early, preventive, and equitable management, improving quality of life for a rapidly aging global population. ACKNOWLEDGMENTS The authors used Grammarly and Copilot for language editing (grammar and readability). CONFLICTS OF INTEREST The authors declare no conflicts of interest. DATA AVAILABILITY Data sharing is not applicable to this article, as no new datasets were generated or analyzed. REFERENCES [1] T. M. Rutkowski, T. Komendziński, and M. Otake-Matsuura, ‘Mild cognitive impairment prediction and cognitive score regression in the 16VOLUME 11, 2023 elderly using eeg topological data analysis and machine learning with awareness assessed in affective reminiscent paradigm,’ Frontiers in Aging Neuroscience, vol. 15, p. 1294139, 2024. [Online]. Available: https://doi.org/10.3389/fnagi.2023.1294139 [2] A. M. Alvi, S. Siuly, and H. Wang, ‘A long short-term memory based framework for early detection of mild cognitive impairment from eeg signals,’ IEEE Transactions on Emerging Topics in Computational In- telligence, vol. 7, no. 2, p. 375–388, April 2023. [3] Z. Li, M. Wu, C. Yin, Z. Wang, J. Wang, L. Chen, and W. Zhao, ‘Machine learning based on the eeg and structural mri can predict different stages of vascular cognitive impairment,’ Frontiers in Aging Neuroscience, vol. 16, p. 1364808, Apr 2024. [Online]. Available: https://doi.org/10.3389/fnagi.2024.1364808 [4] T. J. Alahmadi, A. U. Rahman, Z. A. Alhababi, S. Ali, and H. K. Alkah- tani, ‘Prediction of mild cognitive impairment using eeg signal and bilstm network,’ Machine Learning: Science and Technology, vol. 5, no. 2, p. 025028, 2024. [5] M. Asif, P. Choudhary, and A. A. Dandawate, ‘Computational neuroscience: Recent advancement,’ in Synaptic Plasticity in Neurodegenerative Disorders. CRC Press, 2024, p. 159–190. [Online]. Available: https://doi.org/10.1201/9781003464648-10 [6] C. H. van Dyck, C. J. Swanson, P. Aisen, R. J. Bateman, C. Chen, M. Gee, M. Kanekiyo, D. Li, L. Reyderman, S. Cohen, L. Froelich, S. Katayama, M. Sabbagh, B. Vellas, D. Watson, S. Dhadda, M. Irizarry, L. D. Kramer, and T. Iwatsubo, ‘Lecanemab in Early Alzheimer’s Disease,’ New Eng- land Journal of Medicine, vol. 388, no. 1, p. 9–21, 2023. [7] Biogen Inc., ‘Biogen to realign resources for alzheimer’s disease fran- chise,’ Press release, January 2024, discontinuation of ADUHELM (aducanumab-avwa); commercial availability through November 2024. [8] J. R. Sims, J. A. Zimmer, C. D. Evans, M. Lu, P. Ardayfio, J. Sparks, A. M. Wessels, S. Shcherbinin, H. Wang, E. S. Monkul Nery, E. C. Collins, P. Solomon, S. Salloway, L. G. Apostolova, O. Hansson, C. Ritchie, D. A. Brooks, M. Mintun, D. M. Skovronsky et al., ‘Donanemab in Early Symptomatic Alzheimer Disease: The TRAILBLAZER-ALZ 2 Randomized Clinical Trial,’ JAMA, vol. 330, no. 6, p. 512–527, 2023. [9] C. R. Jack, Jr., J. S. Andrews, T. G. Beach, T. Buracchio, B. Dunn, A. Graf, O. Hansson, C. Ho, W. Jagust, E. McDade, J. L. Molinuevo, O. C. Okonkwo, L. Pani, M. S. Rafii, P. Scheltens, E. Siemers, H. M. Snyder, R. Sperling, C. E. Teunissen, and M. C. Carrillo, ‘Revised criteria for diagnosis and staging of Alzheimer’s disease: Alzheimer’s Association Workgroup,’ Alzheimer’s & Dementia, vol. 20, no. 8, p. 5143–5169, 2024. [10] G. Livingston, J. Huntley, K. Y. Liu et al., ‘Dementia prevention, inter- vention, and care: 2024 report of the Lancet standing Commission,’ The Lancet, vol. 404, no. 10452, p. 572–628, 2024. [11] W.-B. Jiang, L.-M. Zhao, and B.-L. Lu, ‘Large Brain Model for Learning Generic Representations with Tremendous EEG Data in BCI,’ in International Conference on Learning Representations (ICLR), 2024. [Online]. Available: https://arxiv.org/abs/2405.18765 [12] J. Wang, S. Zhao, Z. Luo et al., ‘CBraMod: A Criss-Cross Brain Foundation Model for EEG Decoding,’ in International Conference on Learning Representations (ICLR), 2025. [Online]. Available: https: //arxiv.org/abs/2412.07236 [13] U.S. Food and Drug Administration, ‘FDA Clears First Blood Test Used in Diagnosing Alzheimer’s Disease,’ FDA News Release, May 2025, 510(k) clearance of the Fujirebio Lumipulse G pTau217/β-Amyloid 1-42 Plasma Ratio. [Online]. Available:https://w.fda.gov/news-events/press-announcements/ fda-clears-first-blood-test-used-diagnosing-alzheimers-disease [14] L. D. Baker, M. A. Espeland, R. A. Whitmer, H. M. Snyder, X. Leng, L. Lovato, K. V. Papp, M. Yu, M. Kivipelto, A. S. Alexander, S. An- tkowiak, M. Cleveland, C. Day, R. Elbein, S. Tomaszewski Farias, D. Fel- ton, K. R. Garcia, D. R. Gitelman, S. Graef, M. Howard, J. Katula, K. Lambert, O. Matongo, A. M. McDonald, V. Pavlik, R. Raman, S. Sal- loway, C. Tangney, J. Ventrelle, S. Wilmoth, B. J. Williams, R. Wing, N. Woolard, and M. C. Carrillo, ‘Structured vs Self-Guided Multidomain Lifestyle Interventions for Global Cognitive Function: The US POINTER Randomized Clinical Trial,’ JAMA, vol. 334, no. 8, p. 681–691, 2025, PMID: 40720610. [15] W. Song, W. Wu, Y. Zhao, H. Xu, G. Chen, S. Jin, J. Chen, S. Xian, and J. Liang, ‘Evidence from a meta-analysis and systematic review reveals the global prevalence of mild cognitive impairment,’ Frontiers in Aging Neuroscience, vol. 15, p. 1227112, 2023. [Online]. Available: https://doi.org/10.3389/fnagi.2023.1227112 [16] N. Salari, F. Lotfi, A. Abdolmaleki et al., ‘The global prevalence of mild cognitive impairment in geriatric population with emphasis on influential factors: a systematic review and meta-analysis,’ BMC Geriatrics, vol. 25, p. 313, 2025. [Online]. Available: https://doi.org/10. 1186/s12877-025-05967-w [17] A. J. Mitchell, ‘A meta-analysis of the accuracy of the mini-mental state examination in the detection of dementia and mild cognitive impairment,’ Journal of Psychiatric Research, vol. 43, no. 4, p. 411–431, 2009. [18] E. R. Paitel, C. B. D. Otteman, M. C. Polking, H. J. Licht, and K. A. Nielson, ‘Functional and effective eeg connectivity patterns in alzheimer’s disease and mild cognitive impairment: a systematic review,’ Frontiers in Aging Neuroscience, vol. 17, p. 1496235, Feb 2025. [Online]. Available: https://doi.org/10.3389/fnagi.2025.1496235 [19] M. Mohamed, N. Mohamed, and J. G. Kim, ‘P300 latency with memory performance: A promising biomarker for preclinical stages of alzheimer’s disease,’ Biosensors (Basel), vol. 14, no. 12, p. 616, Dec 2024. [Online]. Available: https://doi.org/10.3390/bios14120616 [20] Zia-Ur-Rehman, M. K. Awang, G. Ali, and M. Faheem, ‘Deep learning techniques for alzheimer’s disease detection in 3d imaging: A systematic review,’ Health Science Reports, vol. 7, no. 9, p. e70025, Sep 2024. [Online]. Available: https://doi.org/10.1002/hsr2.70025 [21] A. Said and H. Göker, ‘Spectral analysis and bi-lstm deep network- based approach in detection of mild cognitive impairment from electroencephalography signals,’ Cognitive Neurodynamics, vol. 18, no. 2, p. 597–614, Apr 2024. [Online]. Available: https://doi.org/10. 1007/s11571-023-10010-y [22] T. Lakhtakia, A. Bondre, P. K. Chand, N. Chaturvedi, S. Choudhary, D. Currey, S. Dutt, A. Khan, M. Kumar, S. Gupta, S. Nagendra, P. V. Reddy, A. Rozatkar, L. Scheuer, Y. Sen, R. Shrivastava, R. Singh, J. Thirthalli, D. K. Tugnawat, A. Bhan, J. A. Naslund, V. Patel, M. Keshavan, U. M. Mehta, and J. Torous, ‘Smartphone digital phenotyping, surveys, and cognitive assessments for global mental health: Initial data and clinical correlations from an international first episode psychosis study,’ Digital Health, vol. 8, p. 20552076221133758, Nov 2022. [Online]. Available: https://doi.org/10.1177/20552076221133758 [23] K. D. Rudd, K. Lawler, M. L. Callisaya, A. D. Bindoff, S. Chiranakorn- Costa, R. Li, J. S. McDonald, K. Salmon, A. J. Noyce, J. C. Vickers, and J. Alty, ‘Hand motor dysfunction is associated with both subjective and objective cognitive impairment across the dementia continuum,’ Dementia and Geriatric Cognitive Disorders, vol. 54, no. 1, p. 10–20, 07 2024. [Online]. Available: https://doi.org/10.1159/000540412 [24] A. Cejudo, M. Arrojo, C. Martín, and A. Almeida, ‘AI and Wearables for Early Detection of Cognitive Impairment and Dementia: Systematic Review,’ Journal of Medical Internet Research, vol. 28, p. e86262, 2026. [25] M. Yan, H. Yin, Q. Meng, S. Wang, Y. Ding, G. Li, C. Wang, and L. Chen, ‘A virtual supermarket program for the screening of mild cognitive impairment in older adults: Diagnostic accuracy study,’ JMIR Serious Games, vol. 9, no. 4, p. e30919, Dec 2021. [26] B. Gómez-Cáceres, I. Cano-López, M. Aliño, and S. Puig-Perez, ‘Effectiveness of virtual reality-based neuropsychological interventions in improving cognitive functioning in patients with mild cognitive impairment: A systematic review and meta-analysis,’ The Clinical Neuropsychologist, vol. 37, no. 7, p. 1337–1370, 2023, PMID: 36416175. [Online]. Available: https://doi.org/10.1080/13854046.2022. 2148283 [27] J. Park, K. Chung, Y. Oh, K. J. Kim, C. O. Kim, and J. Y. Park, ‘Effect of home-based transcranial direct current stimulation on cognitive function in patients with mild cognitive impairment: A two-week intervention,’ Yonsei Medical Journal, vol. 65, no. 6, p. 341–347, Jun 2024. [Online]. Available: https://doi.org/10.3349/ymj.2023.0430 [28] R. Nardone, L. Sebastianelli, V. Versace, D. Ferrazzoli, L. Saltuari, and E. Trinka, ‘Tms–eeg co-registration in patients with mild cognitive impairment, alzheimer’s disease and other dementias: A systematic review,’ Brain Sciences, vol. 11, no. 3, p. 303, 2021. [Online]. Available: https://doi.org/10.3390/brainsci11030303 [29] X. Du, J. Novoa-Laurentiev, J. M. Plasek, Y. Chuang, L. Wang, G. A. Marshall, S. K. Mueller, F. Chang, S. Datta, H. Paek, B. Lin, Q. Wei, X. Wang, J. Wang, H. Ding, F. J. Manion, J. Du, D. W. Bates, and L. Zhou, ‘Enhancing early detection of cognitive decline in the elderly: A comparative study utilizing large language models in clinical notes,’ EBioMedicine, vol. 109, p. 105401, 2024. [Online]. Available: https://doi.org/10.1016/j.ebiom.2024.105401 [30] E. Colita, V. O. Mateescu, D.-G. Olaru, and A. Popa-Wagner, ‘Cognitive decline in ageing and disease: Risk factors, genetics and treatments,’ VOLUME 11, 202317 Current Health Sciences Journal, vol. 50, no. 2, p. 170–180, 2024. [Online]. Available: https://doi.org/10.12865/CHSJ.50.02.02 [31] J. Bae, M. Choi, J. J. Lee, K. H. Lee, and J. U. Kim, ‘Connectivity changes in two-channel prefrontal erp associated with early cognitive decline in the elderly population: Beta band responses to the auditory oddball stimuli,’ Frontiers in Aging Neuroscience, vol. 16, p. 1456169, 2024. [Online]. Available: https://doi.org/10.3389/fnagi.2024.1456169 [32] D. Beltrami, G. Gagliardi, R. Rossini Favretti, E. Ghidoni, F. Tamburini, and L. Calzà, ‘Speech analysis by natural language processing techniques: A possible tool for very early detection of cognitive decline?’ Frontiers in Aging Neuroscience, vol. 10, p. 414837, 2018. [Online]. Available: https://doi.org/10.3389/fnagi.2018.00369 [33] D. Jiang, L. Yan, and F. Mayrand, ‘Emotion expressions and cognitive impairments in the elderly: Review of the contactless detection approach,’ Frontiers in Digital Health, vol. 6, p. 1335289, 2024. [Online]. Available: https://doi.org/10.3389/fdgth.2024.1335289 [34] T. Yamagami, M. Yagi, S. Tanaka, S. Anzai, T. Ueda, Y. Omori, C. Tanaka, and Y. Shiba, ‘Relationship between cognitive decline and daily life gait among elderly people living in the community: A preliminary report,’ Dementia and Geriatric Cognitive Disorders Extra, vol. 13, no. 1, p. 1–9, Dec 2023. [Online]. Available: https://doi.org/10.1159/000528507 [35] H. Chen, Y. Deng, X. Li et al., ‘Factors associated with dementia risk reduction lifestyle in mild cognitive impairment: A cross-sectional study of individuals and their family caregivers,’ BMC Neurology, vol. 25, p. 169, 2025. [Online]. Available: https://doi.org/10.1186/ s12883-025-04183-8 [36] A. Berg, S. Sinclair, A. Acosta-Parra, A. N. Glosson, C. Moreno, and E. Fletcher, ‘Lifestyle factors’ influence on episodic memory: A gradient boosted tree analysis,’ Alzheimer’s & Dementia, vol. 20, p. e095791, 2024. [Online]. Available: https://doi.org/10.1002/alz.095791 [37] S. Y. Jeon and J. L. Kim, ‘Caregiving for a spouse with cognitive impairment: Effects on nutrition and other lifestyle factors,’ Journal of Alzheimer’s Disease, 2021. [Online]. Available: https://doi.org/10.3233/ JAD-210694 [38] H. Sigmundsson, B. H. Dybendal, and S. Grassini, ‘Motion, relation, and passion in brain physiological and cognitive aging,’ Brain Sciences, vol. 12, no. 9, p. 1122, 2022. [Online]. Available: https://doi.org/10.3390/ brainsci12091122 [39] J. K. Burton, D. J. Stott, R. McShane, A. H. Noel-Storr, R. S. Swann- Price, and T. J. Quinn, ‘Informant questionnaire on cognitive decline in the elderly (iqcode) for the early detection of dementia across a variety of healthcare settings,’ Cochrane Database of Systematic Reviews, no. 7, 2021. [Online]. Available: https://doi.org/10.1002/14651858.CD011333. pub3 [40] A. A. Tahami Monfared, W. Ye, A. Sardesai et al., ‘A path to improved alzheimer’s care: Simulating long-term health outcomes of lecanemab in early alzheimer’s disease from the clarity ad trial,’ Neurology and Therapy, vol. 12, p. 863–881, 2023. [Online]. Available: https://doi.org/10.1007/s40120-023-00473-w [41] Z. Ye, ‘Factors influencing memory decline in older adults: A compre- hensive review,’ Studies in Psychological Science, vol. 1, no. 1, p. 27– 41, 2023. [42] M. Acharya, R. C. Deo, X. Tao et al., ‘Deep learning techniques for automated Alzheimer’s and mild cognitive impairment disease using EEG signals: A comprehensive review of the last decade (2013–2024),’ Computer Methods and Programs in Biomedicine, vol. 259, p. 108506, 2025. [43] Z. Yu, A. Mulholland, T. Huang, and Q. Liu, ‘Multimodal AI for Alzheimer Disease Diagnosis: Systematic Review of Datasets, Mod- els, and Modalities,’ Journal of Medical Internet Research, vol. 28, p. e85414, 2026. [44] P. Mobtahej et al., ‘Transformer-Based Deep Learning Approaches for Speech-Based Dementia Detection: A Systematic Review,’ IEEE Journal of Biomedical and Health Informatics, vol. 30, no. 3, p. 2034–2048, 2026. [45] Å. Gausemel and P. Filkuková, ‘Innovations in dementia screening: a systematic review and meta-analysis of virtual reality assessments,’ Frontiers in Psychology, vol. 16, p. 1606562, 2025. [46] S. E. Polk, F. Öhman, J. Hassenstab et al., ‘A scoping review of remote and unsupervised digital cognitive assessments in preclinical Alzheimer’s disease,’ npj Digital Medicine, vol. 8, p. 266, 2025. [47] X. Xin, Q. Liu, S. Jia, S. Li, P. Wang, X. Wang, and X. Wang, ‘Correlation of muscle strength, information processing speed and cognitive function in the elderly with cognitive impairment—evidence from eeg,’ Frontiers in Aging Neuroscience, vol. 17, p. 1496725, 2025. [Online]. Available: https://doi.org/10.3389/fnagi.2025.1496725 [48] W. S. Kumar and S. Ray, ‘Healthy ageing and cognitive impairment alter eeg functional connectivity in distinct frequency bands,’ European Journal of Neuroscience, vol. 58, no. 6, p. 3432–3449, 2023. [Online]. Available: https://doi.org/10.1111/ejn.16114 [49] B. Tóth, B. File, R. Boha, Z. Kardos, Z. Hidasi, Z. A. Gaál, Éva Csibri, P. Salacz, C. J. Stam, and M. Molnár, ‘Eeg network connectivity changes in mild cognitive impairment — preliminary results,’ International Journal of Psychophysiology, vol. 92, no. 1, p. 1–7, 2014. [Online]. Available: https://w.sciencedirect.com/science/ article/pii/S0167876014000403 [50] M. jae Kim, Y. C. Youn, and J. Paik, ‘Deep learning-based eeg analysis to classify normal, mild cognitive impairment, and dementia: Algorithms and dataset,’ NeuroImage, vol. 272, p. 120054, 2023. [Online]. Available: https://w.sciencedirect.com/science/article/pii/S1053811923002008 [51] M. N. A. Tawhid, S. Siuly, E. Kabir, and Y. Li, ‘Exploring frequency band-based biomarkers of eeg signals for mild cognitive impairment detection,’ IEEE Transactions on Neural Systems and Rehabilitation Engineering, vol. 32, p. 189–199, 2024. [52] Q. Chen, W. Hou, X. Wang, W. Zheng, M. Hou, H. Zeng, and L. Cheng, ‘Early warning and screening of elderly cognitive impairment based on machine learning algorithm,’ in 2023 IEEE 3rd International Conference on Computer Communication and Artificial Intelligence (CCAI), May 2023, p. 7–14. [53] A. H. Meghdadi, D. Salat, J. Hamilton, Y. Hong, B. F. Boeve et al., ‘Eeg and erp biosignatures of mild cognitive impairment for longitudinal monitoring of early cognitive decline in alzheimer’s disease,’ PLOS ONE, vol. 19, no. 8, p. e0308137, 2024. [Online]. Available: https://doi.org/10.1371/journal.pone.0308137 [54] Z. Khan, A. Saif, N. Chaudhry, and A. Parveen, ‘Effect of aerobic exercise training on eeg: event-related potential and neuropsychological functions in depressed elderly with mild cognitive impairment,’ Dementia & Neuropsychologia, vol. 17, p. e20220082, 2023. [Online]. Available: https://doi.org/10.1590/1980-5764-DN-2022-0082 [55] G. Praveena and G. Ramesh, ‘Early detection of alzheimer’s disease and dementia using deep convolutional neural networks,’ in 2024 Third In- ternational Conference on Distributed Computing and Electrical Circuits and Electronics (ICDCECE), April 2024, p. 1–5. [56] T. Chattopadhyay, N. A. Joshy, S. S. Ozarkar, K. Buwa, Y. Feng, E. Laltoo, S. I. Thomopoulos, J. E. Villalon, H. Joshi, G. Venkatasubramanian, J. P. John, and P. M. Thompson, ‘Brain age analysis and dementia classifica- tion using convolutional neural networks trained on diffusion mri: Tests in indian and north american cohorts,’ in 2024 46th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), July 2024, p. 1–7. [57] M. Wu, M. Yu, S. Jing, P.-T. Yap, Z. Zhang, and M. Liu, ‘Unpaired volumetric harmonization of brain MRI with conditional latent diffusion,’ Medical Image Analysis, vol. 107, p. 103849, 2026. [58] B. Haché, V. Roca, G. Kuchcinski et al., ‘NeuroHarm-Kit: an open- source toolbox for benchmarking deep-learning harmonization of multi- site T1-weighted MRI,’ NeuroImage, vol. 338, p. 122089, 2026. [59] D. Singh, A. Grazia, M. Dyrba, and S. Teipel, ‘An explainable framework for convolutional neural networks detecting dementia in mri scans,’ Alzheimer’s & Dementia, vol. 20, p. e086103, 2024. [Online]. Available: https://doi.org/10.1002/alz.086103 [60] X. Chen, Y. Li, R. Li, X. Yuan, M. Liu, W. Zhang, and Y. Li, ‘Multiple cross-frequency coupling analysis of resting-state eeg in patients with mild cognitive impairment and alzheimer’s disease,’ Frontiers in Aging Neuroscience, vol. 15, p. 1142085, 2023. [Online]. Available: https://doi.org/10.3389/fnagi.2023.1142085 [61] E. Sibilano, A. Brunetti, D. Buongiorno, M. Lassi, A. Grippo, V. Bessi, S. Micera, A. Mazzoni, and V. Bevilacqua, ‘An attention-based deep learning approach for the classification of subjective cognitive decline and mild cognitive impairment using resting-state eeg,’ Journal of Neural Engineering, vol. 20, no. 1, p. 016048, feb 2023. [Online]. Available: https://dx.doi.org/10.1088/1741-2552/acb96e [62] Y. Liang, P. Li, Y. Wang et al., ‘Application of resting-state EEG theta/alpha power ratio analysis for diagnosing amnestic mild cognitive impairment,’ Scientific Reports, vol. 16, p. 21471, 2026. [63] P. Champetier, C. Albero, F. Raposo Pereira et al., ‘Sleep-like slow waves during resting-state: a promising EEG biomarker of amyloid and neurodegeneration in preclinical Alzheimer’s disease,’ Alzheimer’s & Dementia, vol. 22, no. 6, p. e71514, 2026. 18VOLUME 11, 2023 [64] K. Tripanpitak, A. Wolf, M. Kapitonova et al., ‘Biomarkers for Alzheimer’s Disease and Mild Cognitive Impairment: Recent Advances in Task-Based EEG,’ Journal of Alzheimer’s Disease, vol. 110, no. 1, p. 58–73, 2026. [65] M. Fang, Y. Yan, W. Song et al., ‘Effects of 40-Hz transcranial alternat- ing current stimulation on cognition and neural markers in Alzheimer’s disease: a randomized, sham-controlled trial,’ Alzheimer’s Research & Therapy, vol. 18, no. 1, p. 115, 2026. [66] C. Berezuk, M. Khan, B. L. Callahan, J. Ramirez, S. E. Black, and K. K. Zakzanis, ‘Sex differences in risk factors that predict progression from mild cognitive impairment to alzheimer’s dementia,’ Journal of the International Neuropsychological Society, vol. 29, no. 4, p. 360–368, 2023. [67] H.-J. Kim, J.-H. Lee, E.-n. Cheong, S.-E. Chung, S. Jo, W.-H. Shim, and Y. J. Hong, ‘Elucidating the risk factors for progression from amyloid-negative amnestic mild cognitive impairment to dementia,’ Current Alzheimer Research, vol. 17, no. 10, p. 893–903, 2020. [Online]. Available: https://w.eurekaselect.com/article/111935 [68] H. Joshi, S. Bharath, R. Balachandar, S. Sadanand, H. V. Vishwakarma, S. Aiyappan, J. Saini, K. J. Kumar, J. P. John, and M. Varghese, ‘Differentiation of early alzheimer’s disease, mild cognitive impairment, and cognitively healthy elderly samples using multimodal neuroimaging indices,’ Brain Connectivity, vol. 9, no. 9, p. 730–741, 2019, PMID: 31389245. [Online]. Available: https://doi.org/10.1089/brain.2019.0676 [69] P. Li, Q. Huang, S. Ban, Y. Qiao, J. Wu, Y. Zhai, X. Du, F. Hua, and J. Su, ‘Altered default mode network is associated with cognitive impairment in cadasil as revealed by multimodal neuroimaging,’ Frontiers in Neurology, vol. 12, p. 735033, 2021. [Online]. Available: https://doi.org/10.3389/fneur.2021.735033 [70] S. Kim, S.-M. Wang, D. W. Kang, Y. H. Um, H. M. Yoon, S. Lee, Y. S. Choe, R. E. Kim, D. Kim, C. U. Lee, and H. K. Lim, ‘Development of a prediction model for cognitive impairment of sarcopenia using multimodal neuroimaging in non-demented older adults,’ Alzheimer’s & Dementia, vol. 20, no. 7, p. 4868–4878, 2024. [Online]. Available: https://alz-journals.onlinelibrary.wiley.com/doi/abs/10.1002/alz.14054 [71] S. Sun, D. Liu, Y. Zhou, G. Yang, L. Cui, X. Xu, Y. Guo, T. Sun, J. Jiang, N. Li, Y. Wang, S. Li, X. Wang, L. Fan, and F. Cao, ‘Longitudinal real world correlation study of blood pressure and novel features of cerebral magnetic resonance angiography by artificial intelligence analysis on elderly cognitive impairment,’ Frontiers in Aging Neuroscience, vol. 15, p. 1121152, 2023. [Online]. Available: https://doi.org/10.3389/fnagi.2023.1121152 [72] S. A. Graham, E. E. Lee, D. V. Jeste, R. V. Patten, E. W. Twamley, C. Nebeker, Y. Yamada, C. Kim, and C. A. Depp, ‘Artificial intelligence approaches to predicting and detecting cognitive decline in older adults: A conceptual review,’ Psychiatry Research, vol. 284, p. 112732, 2019. [Online]. Available: https://doi.org/10.1016/j.psychres.2019.112732 [73] L. Falaschetti, G. Biagetti, M. Alessandrini, C. Turchetti, S. Luzzi, and P. Crippa, ‘Multi-class detection of neurodegenerative diseases from eeg signals using lightweight lstm neural networks,’ Sensors, vol. 24, no. 20, p. 6721, 2023. [Online]. Available: https://doi.org/10.3390/s24206721 [74] M. Khosravi et al., ‘Fusing convolutional learning and attention-based bi-lstm networks for early alzheimer’s diagnosis from eeg signals towards iomt,’ Scientific Reports, vol. 14, no. 1, p. 26002, Oct 2024. [Online]. Available: https://doi.org/10.1038/s41598-024-77876-8 [75] M. Asif, M. T. Vinodbhai, S. Mishra, A. Gupta, and U. S. Tiwary, ‘Emotion recognition in VAD space during emotional events using CNN-GRU hybrid model on EEG signals,’ in Intelligent Human Computer Interaction (IHCI 2022), ser. Lecture Notes in Computer Science, vol. 13741. Springer, 2023, p. 75–84. [Online]. Available: https://doi.org/10.1007/978-3-031-27199-1_8 [76] M. Asif, N. Ali, S. Mishra, A. Dandawate, and U. S. Tiwary, ‘Deep fuzzy framework for emotion recognition using EEG signals and emotion representation in type-2 fuzzy VAD space,’ arXiv preprint arXiv:2401.07892, 2024. [Online]. Available: https://doi.org/10.48550/ arXiv.2401.07892 [77] B. A. C. Ramalho, L. R. Bortolato, N. D. Gomes, L. Wichert-Ana, F. E. Padovan-Neto, M. A. A. da Silva, and K. J. C. C. de Lacerda, ‘The impact of the orientation of mri slices on the accuracy of alzheimer’s disease classification using convolutional neural networks (cnns),’ Journal of Medical Artificial Intelligence, vol. 7, no. 0, 2024. [Online]. Available: https://jmai.amegroups.org/article/view/9069 [78] A. M. El-Assy, H. M. Amer, H. M. Ibrahim, and M. A. Mohamed, ‘A novel cnn architecture for accurate early detection and classification of alzheimer’s disease using mri data,’ Scientific Reports, vol. 14, no. 1, p. 1–19, 2024. [Online]. Available: https://doi.org/10.1038/ s41598-024-53733-6 [79] M. U. Ali, K. S. Kim, M. Khalid, M. Farrash, A. Zafar, and S. W. Lee, ‘Enhancing alzheimer’s disease diagnosis and staging: A multistage cnn framework using mri,’ Frontiers in Psychiatry, vol. 15, p. 1395563, 2024. [Online]. Available: https://doi.org/10.3389/fpsyt.2024.1395563 [80] S. Sinha, S. I. Thomopoulos, P. Lam, A. Muir, and P. M. Thompson, ‘Alzheimer’s disease classification accuracy is improved by mri harmonization based on attention-guided generative adversarial networks,’ in Proceedings of the 17th International Symposium on Medical Information Processing and Analysis, ser. Proc. SPIE, vol. 12088, December 2021, p. 120880L. [Online]. Available: https://doi.org/10.1117/12.2606155 [81] M. Yao, J. Liu, Y. Pu, and K. H. Katie Chan, ‘Multi-class prediction of cognitively normal / mild cognitive impairment / alzheimer’s disease status in dementia based on convolutional neural networks with attention mechanism,’ in 2024 46th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), 2024, p. 1–7. [82] W. Ibrar, M. A. Khan, A. Hamza et al., ‘A novel interpreted deep network for Alzheimer’s disease prediction based on inverted self attention and vision transformer,’ Scientific Reports, vol. 15, p. 29974, 2025. [83] S. A. Martin, A. Zhao, J. Qu et al., ‘Investigating the Utility of Explain- able Artificial Intelligence for Neuroimaging-Based Dementia Diagnosis and Prognosis,’ Human Brain Mapping, vol. 47, no. 2, p. e70456, 2026. [84] M. Asif, A. Gupta, A. Aditya, S. Mishra, and U. S. Tiwary, ‘Brain multi- region information fusion using attentional transformer for EEG based affective computing,’ in 2023 IEEE 20th India Council International Conference (INDICON). Hyderabad, India: IEEE, 2023, p. 771– 775. [Online]. Available: https://doi.org/10.1109/INDICON59947.2023. 10440791 [85] J.-E. Ding, A. Zilverstand, S. Yang, A. C.-C. Yang, and F. Liu, ‘Variational Mixture of Graph Neural Experts for Alzheimer’s Disease Recognition across Frequency Bands in EEG Brain Networks,’ arXiv preprint arXiv:2510.11917, 2025, preprint — not peer reviewed. [Online]. Available: https://arxiv.org/abs/2510.11917 [86] Y. Wang, N. Huang, N. Mammone, M. Cecchi, and X. Zhang, ‘LEAD: An EEG Foundation Model for Alzheimer’s Disease Detection,’ arXiv preprint arXiv:2502.01678, 2025, preprint — not peer reviewed. [Online]. Available: https://arxiv.org/abs/2502.01678 [87] A. Miltiadous, A. Ntetska, V. Aspiotis et al., ‘The AHEPA EEG bench- mark: setting the standard for machine learning in dementia diagnosis, a scoping review,’ Cognitive Neurodynamics, vol. 20, no. 1, p. 95, 2026. [88] M. Zare, ‘Stress-Testing EEG Foundation Models for Clinical Decoding: Dataset Identity and Targeted Negative Controls,’ arXiv preprint arXiv:2607.24519, 2026, preprint — not peer reviewed (posted 27 Jul 2026, single author). [Online]. Available: https://arxiv.org/abs/2607. 24519 [89] E. Zendehrouh, M. S. E. Sendi, A. Abrol, I. Batta, R. Hassanzadeh, and V. D. Calhoun, ‘Towards a multimodal neuroimaging-based risk score for mild cognitive impairment by combining clinical studies with a large (n>37000) population-based study,’ medRxiv, March 2024, preprint. [Online]. Available: https://doi.org/10.1101/2024.03.12.24303873 [90] V. Adarsh, G. R. Gangadharan, U. Fiore, and P. Zanetti, ‘Multimodal classification of alzheimer’s disease and mild cognitive impairment using custom mkscddl kernel over cnn with transparent decision-making for explainable diagnosis,’ Scientific Reports, vol. 14, no. 1, p. 1–16, 2024. [Online]. Available: https://doi.org/10.1038/s41598-024-52185-2 [91] Y. Zhu, F. Wang, P. Ning, Y. Zhu, L. Zhang, K. Li, B. Liu, H. Ren, Z. Xu, A. Pang, and X. Yang, ‘Multimodal neuroimaging-based prediction of parkinson’s disease with mild cognitive impairment using machine learning technique,’ NPJ Parkinson’s Disease, vol. 10, no. 1, p. 1–11, 2024. [Online]. Available: https://doi.org/10.1038/s41531-024-00828-6 [92] M. N. Sabbagh, M. Boada, S. Borson, M. Chilukuri, P. Doraiswamy, B. Dubois, J. Ingram, A. Iwata, A. Porsteinsson, K. Possin, G. Rabinovici, B. Vellas, S. Chao, A. Vergallo, and H. Hampel, ‘Rationale for early diagnosis of mild cognitive impairment (mci) supported by emerging digital technologies,’ The Journal of Prevention of Alzheimer’s Disease, vol. 7, no. 3, p. 158–164, 2020. [Online]. Available: https://doi.org/10.14283/jpad.2020.19 [93] Z. Dong, H. Liu, X. Ge et al., ‘Brain Region-Centered MultiModal Hypergraph Fusion for MCI Conversion Prediction,’ IEEE Journal of Biomedical and Health Informatics, 2026, early access. VOLUME 11, 202319 [94] K. Kiguchi, Y. Tu, K. Ajito, F. Alnajjar, and K. Murase, ‘Multi-modal Integration Analysis of Alzheimer’s Disease Using Large Language Mod- els and Knowledge Graphs,’ IEEE Access, vol. 13, p. 113 718–113 735, 2025. [95] M. Kallel, B. Park, K. Seo, and S.-E. Kim, ‘Multimodal machine learning model for mci detection using eeg, mri and vr data,’ in 2024 International Technical Conference on Circuits/Systems, Computers, and Communica- tions (ITC-CSCC), 2024, p. 1–6. [96] M. H. Modarres, C. Kalafatis, P. Apostolou, and N. Tabet, ‘The use of the integrated cognitive assessment to improve the efficiency of primary care referrals to memory services in the accelerating dementia pathway technologies study,’ Frontiers in Aging Neuroscience, vol. 15, p. 1243316, 2023. [Online]. Available: https://doi.org/10.3389/fnagi.2023. 1243316 [97] C. Kalafatis, M. Modarres, P. Apostolou, N. Tabet, and S.-M. Khaligh- Razavi, ‘The use of a computerized cognitive assessment to improve the efficiency of primary care referrals to memory services: Protocol for the accelerating dementia pathway technologies (adept) study,’ JMIR Research Protocols, vol. 11, no. 1, p. e34475, 2022. [Online]. Available: https://w.researchprotocols.org/2022/1/e34475 [98] M. R. Lima, A. Capstick, F. Geranmayeh et al., ‘Evaluating spoken lan- guage as a biomarker for automated screening of cognitive impairment,’ Communications Medicine, vol. 6, p. 6, 2025. [99] A. Zolnour et al., ‘LLMCARE: early detection of cognitive impairment via transformer models enhanced by LLM-generated synthetic data,’ Frontiers in Artificial Intelligence, vol. 8, p. 1669896, 2025. [100] R. Shankar, Z. Goh, F. Devi, and Q. Xu, ‘A systematic review of explain- able artificial intelligence methods for speech-based cognitive decline detection,’ npj Digital Medicine, vol. 8, p. 724, 2025. [101] S. Cavedoni, A. Chirico, E. Pedroli, P. Cipresso, and G. Riva, ‘Digital biomarkers for the early detection of mild cognitive impairment: Artificial intelligence meets virtual reality,’ Frontiers in Human Neuroscience, vol. 14, p. 535098, 2020. [Online]. Available: https: //doi.org/10.3389/fnhum.2020.00245 [102] E. Gkintoni, S. P. Vassilopoulos, G. Nikolaou, and A. Vantarakis, ‘Neurotechnological approaches to cognitive rehabilitation in mild cognitive impairment: A systematic review of neuromodulation, eeg, virtual reality, and emerging ai applications,’ Brain Sciences, vol. 15, no. 6, p. 582, 2025. [Online]. Available: https://doi.org/10.3390/ brainsci15060582 [103] M. Cai, Y. Zhang, S. Chen, Z. Wu, and L. Zhu, ‘The past, present, and future of research on neuroinflammation-induced mild cognitive impairment: A bibliometric analysis,’ Frontiers in Aging Neuroscience, vol. 14, p. 968444, 2022. [Online]. Available: https: //doi.org/10.3389/fnagi.2022.968444 [104] C. Tuena, S. Serino, C. Stramba-Badiale et al., ‘Egocentric Spatial Mem- ory Deficit in Amnestic Mild Cognitive Impairment Revealed Through Virtual Reality: Cross-Sectional Study,’ JMIR Aging, vol. 9, p. e79224, 2026. [105] F. Nerrise, N. Schütz, Q. Zhao et al., ‘A framework of digital biomarkers for neurodegenerative diseases,’ Nature Reviews Bioengineering, vol. 4, no. 7, p. 675–694, 2026. [106] R. Wu, A. Li, C. Xue, J. Chai, Y. Qiang, J. Zhao, and L. Wang, ‘Screening for mild cognitive impairment with speech interaction based on virtual reality and wearable devices,’ Brain Sciences, vol. 13, no. 8, p. 1222, 2023. [Online]. Available: https://doi.org/10.3390/brainsci13081222 [107] Y. G. Rykov, M. D. Patterson, B. A. Gangwar, S. B. Jabar, J. Leonardo, K. P. Ng, and N. Kandiah, ‘Predicting cognitive scores from wearable- based digital physiological features using machine learning: Data from a clinical trial in mild cognitive impairment,’ BMC Medicine, vol. 22, p. 36, 2024. [Online]. Available: https://doi.org/10.1186/s12916-024-03252-y [108] E. Gleaton, E. Parcell, J. Erickson, and W. B. Fain, ‘Exploring longitudinal physiological monitoring: Insights from utilizing a commercial wearable device among older adults with mild cognitive impairment,’ in Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 2024. [Online]. Available: https://doi.org/10.1177/10711813241272130 [109] P. M. Butler, J. Yang, R. Brown, M. Hobbs, A. Becker, J. Penalver-Andres, P. Syz, S. Muller, G. Cosne, A. Juraver, H. H. Song, P. Saha-Chaudhuri, D. Roggen, A. Scotland, N. Silveira, G. Demircioglu, A. Gabelle, R. Hughes, M. G. Erkkinen, and S. Belachew, ‘Smartwatch- and smartphone-based remote assessment of brain health and detection of mild cognitive impairment,’ Nature Medicine, vol. 31, no. 3, p. 829, 2025. [Online]. Available: https://doi.org/10.1038/s41591-024-03475-9 [110] J. Dieffenderfer, A. Brewer, M. A. Noonan, M. Smith, E. Eichenlaub, K. L. Haley, A. Jacks, E. Lobaton, S. D. Neupert, T. M. Hess, J. R. Franz, S. K. Ghosh, V. Misra, and A. Bozkurt, ‘A wearable system for continuous monitoring and assessment of speech, gait, and cognitive decline for early diagnosis of adrd,’ in 2023 45th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), 2023, p. 1–6. [111] J. Razjouyan, B. Najafi, M. Horstman, A. Sharafkhaneh, M. Amirmazaheri, H. Zhou, M. E. Kunik, and A. Naik, ‘Toward using wearables to remotely monitor cognitive frailty in community- living older adults: An observational study,’ Sensors, vol. 20, no. 8, p. 2218, 2019. [Online]. Available: https://doi.org/10.3390/s20082218 [112] H. Azami, M. Mirjalili, T. K. Rajji, C.-T. Wu, A. Humeau-Heurtier, T.- P. Jung, C.-S. Wei, T.-T. Trinh, and Y.-H. Liu, ‘Electroencephalogram and event-related potential in mild cognitive impairment: Recent develop- ments in signal processing, machine learning, and deep learning,’ IEEE Journal of Selected Areas in Sensors, vol. 2, p. 162–184, 2025. [113] M. Asif, N. Ali, Aditya, Diya, S. Mishra, and U. S. Tiwary, ‘Spatially-infused spectrograms for robust EEG feature extraction in deep learning frameworks,’ in ICASSP 2026 – 2026 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, May 2026, p. 6881–6885. [Online]. Available: https://doi.org/10. 1109/ICASSP55912.2026.11464787 [114] E. M. Ye, H. Sun, P. V. Krishnamurthy, N. Adra, W. Ganglberger, R. J. Thomas, A. D. Lam, and M. B. Westover, ‘Dementia detection from brain activity during sleep,’ Sleep, vol. 46, no. 3, 2023. [Online]. Available: https://doi.org/10.1093/sleep/zsac286 [115] D. Kim, J. Park, H. Choi, H. Ryu, M. Loeser, and K. Seo, ‘Early detec- tion of patients with mild cognitive impairment through eeg-ssvep-based machine learning model,’ IEEE Access, vol. 12, p. 172 101–172 114, 2024. [116] T. M. Rutkowski, M. S. Abe, H. Sugimoto, and M. Otake-Matsuura, ‘Mild cognitive impairment detection with machine learning and topo- logical data analysis applied to eeg time-series in facial emotion oddball paradigm,’ in 2023 45th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), 2023, p. 1–4. [117] T. M. Rutkowski, Q. Zhao, M. S. Abe, and M. Otake, ‘Ai neurotechnology for aging societies - task-load and dementia eeg digital biomarker development using information geometry machine learning methods,’ ArXiv, vol. abs/1811.12642, 2018. [Online]. Available: https://api. semanticscholar.org/CorpusID:54434893 [118] M. J. Sedghizadeh, H. Aghajan, Z. Vahabi et al., ‘Network synchronization deficits caused by dementia and alzheimer’s disease serve as topographical biomarkers: a pilot study,’ Brain Structure and Function, vol. 227, p. 2957–2969, 2022. [Online]. Available: https://doi.org/10.1007/s00429-022-02554-2 [119] T. M. Rutkowski, M. S. Abe, and M. Otake-Matsuura, ‘Neurotechnology and ai approach for early dementia onset biomarker from eeg in emotional stimulus evaluation task,’ in 2021 43rd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), 2021, p. 6675–6678. [120] T. M. Rutkowski, S. Narębski, and T. Komendziński, ‘Early detection of cognitive impairment in elderly using a passive fpvs-eeg bci and machine learning – extended version,’ arXiv, 2025. [Online]. Available: https://arxiv.org/abs/2504.10973 [121] J. Huang, C. Wang, W. Zhao, A. Grau, X. Xue, and F. Zhang, ‘Ltdnet- eeg: A lightweight network of portable/wearable devices for real-time eeg signal denoising,’ IEEE Transactions on Consumer Electronics, vol. 70, no. 3, p. 5561–5575, 2024. [122] R. Zanetti, A. Arza, A. Aminifar, and D. Atienza, ‘Real-time eeg-based cognitive workload monitoring on wearable devices,’ IEEE Transactions on Biomedical Engineering, vol. 69, no. 1, p. 265–277, 2022. [123] H. Huang, J. Chen, J. Xiao, D. Chen, J. Zhang, J. Pan, and Y. Li, ‘Real- time attention regulation and cognitive monitoring using a wearable eeg- based bci,’ IEEE Transactions on Biomedical Engineering, vol. 72, no. 2, p. 716–724, 2025. [124] Q. Zhao, R. Yang, X. Zhu, C. Li, X. Gao, Y. Li, H. Yu, J. Wang, J. An, Z. Tan, and Z. Zhao, ‘Systematic comparison of signal quality in portable and wearable wireless eeg devices: Methods and standards,’ in Ergonomics In Design, ser. AHFE Open Access, P.-L. Rau, Ed., vol. 129. USA: AHFE International, 2024. [Online]. Available: http://doi.org/10.54941/ahfe1004813 [125] N.-D. Mai, H. Barki, and W.-Y. Chung, ‘A wearable bte-eeg embed- ded device for emotion monitoring with quantum machine learning,’ in 20VOLUME 11, 2023 2024 Tenth International Conference on Communications and Electronics (ICCE), 2024, p. 562–566. [126] J. Olichney, J. Xia, K. J. Church, and H. J. Moebius, ‘Predictive power of cognitive biomarkers in neurodegenerative disease drug development: Utility of the p300 event-related potential,’ Neural Plasticity, vol. 2022, no. 1, p. 2104880, 2021. [Online]. Available: https://doi.org/10.1155/2022/2104880 [127] R. Rodríguez-Labrada, L. Velázquez-Pérez, R. Ortega-Sánchez et al., ‘Insights into cognitive decline in spinocerebellar ataxia type 2: a p300 event-related brain potential study,’ Cerebellum & Ataxias, vol. 6, no. 3, 2019. [Online]. Available: https://doi.org/10.1186/s40673-019-0097-2 [128] J. Xia, A. Mazaheri, K. Segaert, D. P. Salmon, D. Harvey, K. Shapiro, M. Kutas, and J. M. Olichney, ‘Event-related potential and eeg oscillatory predictors of verbal memory in mild cognitive impairment,’ Brain Communications, vol. 2, no. 2, 2020. [Online]. Available: https://doi.org/10.1093/braincomms/fcaa213 [129] K. Chu, W. Lei, M. Wu, J. Fuh, S. Wang, I. T. French, W. Chang, C. Chang, N. E. Huang, W. Liang, and C. Juan, ‘A holo-spectral eeg analysis provides an early detection of cognitive decline and predicts the progression to alzheimer’s disease,’ Frontiers in Aging Neuroscience, vol. 15, p. 1195424, 2023. [Online]. Available: https: //doi.org/10.3389/fnagi.2023.1195424 [130] L. Zhao, Y. Ma, and W. Li, ‘Study on the construction of risk prediction model and efficacy validation of cognitive decline in elderly patients with type 2 diabetes,’ Journal of Men’s Health, vol. 20, no. 3, p. 99–105, 2024. [131] A. S. Dias Portela, V. Saxena, E. Rosenn, H. Wang, S. Masieri, J. Palmieri, and G. M. Pasinetti, ‘Role of artificial intelligence in multinomial decisions and preventative nutrition in alzheimer’s disease,’ Molecular Nutrition & Food Research, vol. 68, no. 13, p. 2300605, 2024. [Online]. Available: https://doi.org/10.1002/mnfr.202300605 [132] M. Gao, J. Wang, Y. Qiu, Y. Chen, Q. Cao, Y. Pan, Y. Cao, S. Han, X. Yan, X. Xu, X. Fang, and F. Lian, ‘Association between dietary diversity and subjective cognitive decline in the middle-aged and elderly chinese population: A cross-sectional study,’ Nutrients, vol. 16, no. 21, p. 3603, 2023. [Online]. Available: https://doi.org/10.3390/nu16213603 [133] S.-J. Lim and J.-H. Park, ‘The study of the convergent factors of the lifestyle on the cognitive decline among elderly,’ Journal of the Korea Convergence Society, vol. 11, no. 8, p. 229–236, 2020. [Online]. Available: https://doi.org/10.15207/JKCS.2020.11.8.229 [134] M. Ahmadzadeh, T. D. Cosco, J. R. Best, G. J. Christie, and S. DiPaola, ‘Predictors of the rate of cognitive decline in older adults using machine learning,’ PLOS ONE, vol. 18, no. 3, p. e0280029, 2023. [Online]. Available: https://doi.org/10.1371/journal.pone.0280029 [135] J. Alcina, L. R. Guevara, A. Duarte, M. Villafana, D. Gonzalez, and Y. P. Haddock, ‘H - 03 modifiable risk factors associated with cognitive decline in hispanic elders: A literature review,’ Archives of Clinical Neuropsychology, vol. 38, no. 7, p. 1484, 2023. [Online]. Available: https://doi.org/10.1093/arclin/acad067.321 [136] M. A. Klados, C. Styliadis, C. A. Frantzidis, E. Paraskevopoulos, and P. D. Bamidis, ‘Beta-band functional connectivity is reorganized in mild cognitive impairment after combined computerized physical and cognitive training,’ Frontiers in Neuroscience, vol. 10, p. 152218, 2016. [Online]. Available: https://doi.org/10.3389/fnins.2016.00055 [137] M. Basta, E. Skourti, C. Alexopoulou, A. Zampetakis, A. Ganiaris, M. Aligizaki, P. Simos, and A. N. Vgontzas, ‘Cretan aging cohort- phase i: Methodology and descriptive characteristics of a long-term longitudinal study on predictors of cognitive decline in non-demented elderly from crete, greece,’ Healthcare, vol. 11, no. 5, p. 703, 2022. [Online]. Available: https://doi.org/10.3390/healthcare11050703 [138] M. Asif, S. Mishra, M. T. Vinodbhai, and U. S. Tiwary, ‘Emotion recognition using temporally localized emotional events in EEG with naturalistic context: DENS# dataset,’ IEEE Access, vol. 11, p. 39 913– 39 925, 2023. [Online]. Available: https://doi.org/10.1109/ACCESS. 2023.3266804 [139] S. Mishra, M. Asif, N. Srinivasan, and U. S. Tiwary, ‘Dataset on emotion with naturalistic stimuli (DENS) on indian samples,’ bioRxiv, 2022, preprint. [Online]. Available: https://doi.org/10.1101/2021.08.04.455041 [140] M. Asif, S. Mishra, A. Sonker, S. Gupta, S. K. Maurya, and U. S. Tiwary, ‘Proactive emotion tracker: AI-driven continuous mood and emotion monitoring,’ arXiv preprint arXiv:2401.13722, 2024. [Online]. Available: https://doi.org/10.48550/arXiv.2401.13722 [141] T. M. Rutkowski, M. S. Abe, S. Tokunaga, T. Komendziński, and M. Otake-Matsuura, ‘Dementia digital neuro-biomarker study from theta-band eeg fluctuation analysis in facial and emotional identification short-term memory oddball paradigm,’ in 2022 44th Annual Interna- tional Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), 2022, p. 4056–4059. [142] K. Dhana, N. T. Aggarwal, K. B. Rajan, L. L. Barnes, D. A. Evans, and M. C. Morris, ‘Apoe4, lifestyle factors, and cognitive decline: A population-based cohort study,’ Alzheimer’s & Dementia, vol. 16, p. e043228, 2020. [Online]. Available: https://doi.org/10.1002/alz.043228 [143] B. Dubois, N. Villain, L. Schneider, N. Fox, N. Campbell, D. Galasko, M. Kivipelto, F. Jessen, B. Winblad, J. Cummings, G. B. Frisoni, R. Bate- man, B. Vellas, G. Wilcock, Z. Ismail, H. Feldman, R. C. Petersen, and P. Scheltens, ‘Alzheimer disease as a clinical-biological construct—an international working group recommendation,’ JAMA Neurology, vol. 81, no. 12, p. 1304–1311, 2024. [144] J. A. Schneider, Z. Arvanitakis, W. Bang, and D. A. Bennett, ‘Mixed brain pathologies account for most dementia cases in community-dwelling older persons,’ Neurology, vol. 69, no. 24, p. 2197–2204, 2007. [145] K. V. Robson, S. Slight, and R. Mc Ardle, ‘Who is at risk of dementia? trajectories of cognitive decline from mild cognitive impairment to dementia and the factors which mediate this decline,’ Alzheimer’s & Dementia, vol. 19, p. e062909, 2023. [Online]. Available: https://doi.org/10.1002/alz.062909 [146] S. Palmqvist, P. Tideman, N. Mattsson-Carlgren, S. E. Schindler, R. Smith, R. Ossenkoppele, S. Calling, T. West, M. Monane, P. B. Vergh- ese, J. B. Braunstein, K. Blennow, S. Janelidze, E. Stomrud, G. Salvadó, and O. Hansson, ‘Blood Biomarkers to Detect Alzheimer Disease in Primary Care and Secondary Care,’ JAMA, vol. 332, no. 15, p. 1245– 1257, 2024. [147] N. R. Barthélemy, G. Salvadó, S. E. Schindler, Y. He, S. Janelidze, L. E. Collij, B. Saef, R. L. Henson, C. D. Chen, B. A. Gordon, Y. Li, R. La Joie, T. L. S. Benzinger, J. C. Morris, N. Mattsson-Carlgren, S. Palmqvist, R. Ossenkoppele, G. D. Rabinovici, E. Stomrud, R. J. Bateman, and O. Hansson, ‘Highly accurate blood test for Alzheimer’s disease is similar or superior to clinical cerebrospinal fluid tests,’ Nature Medicine, vol. 30, no. 4, p. 1085–1095, 2024. [148] N. J. Ashton, W. S. Brum, G. Di Molfetta, A. L. Benedet, B. Arslan, E. Jonaitis, R. E. Langhough, K. Cody, R. Wilson, C. M. Carlsson, E. Vanmechelen, L. Montoliu-Gaya, J. Lantero-Rodriguez, N. Rahmouni, C. Tissot, J. Stevenson, S. Servaes, J. Therriault, T. Pascoal, A. Lleó, D. Alcolea, J. Fortea, P. Rosa-Neto, S. Johnson, A. Jeromin, K. Blennow, and H. Zetterberg, ‘Diagnostic Accuracy of a Plasma Phosphorylated Tau 217 Immunoassay for Alzheimer Disease Pathology,’ JAMA Neurology, vol. 81, no. 3, p. 255–263, 2024. [149] C. Leurent and M. Ehlers, ‘Digital technologies for cognitive assessment to accelerate drug development in alzheimer’s disease,’ Clinical Pharmacology & Therapeutics, vol. 98, no. 5, p. 475–476, 2015. [Online]. Available: https://doi.org/10.1002/cpt.212 [150] K. Niotis, K. Akiyoshi, C. Carlton, and R. Isaacson, ‘Dementia prevention in clinical practice,’ Seminars in Neurology, vol. 42, no. 5, p. 525–548, 2022. [Online]. Available: https://doi.org/10.1055/ s-0042-1759580 [151] S. Li, X. Xie, D. Liu, G. Cheng, F. Hu, D. Zeng, X. Chen, L. Jia, Y. Wang, X. Bu, C. Qiu, F. Gao, J. Gu, M. Liu, Y. Li, Y. Zhou, H. Chang, Y. Ou, L. Xu, and Y. Zeng, ‘China initiative for multi-domain intervention (china-in-mudi) to prevent cognitive decline: Study design and progress,’ The Journal of Prevention of Alzheimer’s Disease, vol. 11, no. 3, p. 589– 600, 2024. [Online]. Available: https://doi.org/10.14283/jpad.2024.63 [152] J. Ulbl and M. Rakusa, ‘The importance of subjective cognitive de- cline recognition and the potential of molecular and neurophysiological biomarkers—a systematic review,’ International Journal of Molecular Sciences, vol. 24, no. 12, p. 10158, 2022. [153] D. M. Cammisuli, R. Ceravolo, and U. Bonuccelli, ‘Non- pharmacological interventions for parkinson’s disease mild cognitive impairment: future directions for research,’ Neural Regeneration Research, vol. 15, no. 9, p. 1650–1651, September 2020. [154] N. Öksüz, R. Ghouri, B. Taşdelen, D. Uludüz, and A. Özge, ‘Mild cognitive impairment progression and alzheimer’s disease risk: A comprehensive analysis of 3553 cases over 203 months,’ Journal of Clinical Medicine, vol. 13, no. 2, p. 518, 2023. [Online]. Available: https://doi.org/10.3390/jcm13020518 [155] P. Iso-Markku, U. M. Kujala, K. Knittle, J. Polet, E. Vuoksimaa, and K. Waller, ‘Physical activity as a protective factor for dementia and alzheimer’s disease: Systematic review, meta-analysis and quality assessment of cohort and case–control studies,’ British Journal of VOLUME 11, 202321 Sports Medicine, vol. 56, no. 12, p. 701, 2022. [Online]. Available: https://doi.org/10.1136/bjsports-2021-104981 [156] B. Singh, A. K. Parsaik, M. M. Mielke, P. J. Erwin, D. S. Knopman, R. C. Petersen, and R. O. Roberts, ‘Association of mediterranean diet with mild cognitive impairment and alzheimer’s disease: A systematic review and meta-analysis,’ Journal of Alzheimer’s Disease, vol. 39, no. 2, p. 271– 282, 2014. [157] J. D. Williamson, N. M. Pajewski, A. P. Auchus et al., ‘Effect of intensive vs standard blood pressure control on probable dementia: A randomized clinical trial,’ JAMA, vol. 321, no. 6, p. 553–561, 2019. [158] B. S. Y. Yeo, H. J. J. Song, E. M. S. Toh, L. S. Ng, C. S. H. Ho, R. Ho, R. A. Merchant, B. K. J. Tan, and W. S. Loh, ‘Association of hearing aids and cochlear implants with cognitive decline and dementia: A systematic review and meta-analysis,’ JAMA Neurology, vol. 80, no. 2, p. 134–141, 2023. [159] S. E. Lakhan, ‘Cognitive resilience for the prevention of mild cognitive impairment in subjective cognitive decline: A monte carlo simulation of a digital therapeutic targeting dementia risk factors,’ Cureus, vol. 17, no. 5, p. e83643, 2025. [Online]. Available: https: //doi.org/10.7759/cureus.83643 [160] L. S. Elias-Sonnenschein, W. Viechtbauer, I. H. G. B. Ramakers, F. R. J. Verhey, and P. J. Visser, ‘Predictive value of APOE-ε4 allele for pro- gression from MCI to AD-type dementia: a meta-analysis,’ Journal of Neurology, Neurosurgery & Psychiatry, vol. 82, no. 10, p. 1149–1156, 2011. [161] M. Kubota, S. Bun, K. Takahata et al., ‘Plasma biomarkers for early detection of alzheimer’s disease: A cross-sectional study in a japanese cohort,’ Alzheimer’s Research & Therapy, vol. 17, p. 131, 2025. [Online]. Available: https://doi.org/10.1186/s13195-025-01778-8 [162] J. Martin, N. Reid, D. D. Ward, S. King, R. E. Hubbard, and E. H. Gordon, ‘Investigating sex differences in risk and protective factors in the progression of mild cognitive impairment to dementia: A systematic review,’ Journal of Alzheimer’s Disease, 2023. [Online]. Available: https://doi.org/10.3233/JAD-230700 [163] W. Yang, F. Guan, L. Yang, G. Shou, F. Zhu, Y. Xu, Y. Meng, M. Li, and W. Dong, ‘Highly sensitive blood-based biomarkers detection of beta-amyloid and phosphorylated-tau181 for alzheimer’s disease,’ Frontiers in Neurology, vol. 15, p. 1445479, 2024. [Online]. Available: https://doi.org/10.3389/fneur.2024.1445479 [164] S. Palmqvist, N. Warmenhoven, F. Anastasi et al., ‘Plasma phospho- tau217 for Alzheimer’s disease diagnosis in primary and secondary care using a fully automated platform,’ Nature Medicine, vol. 31, no. 6, p. 2036–2043, 2025. [165] G. Grande, M. Valletta, D. Rizzuto et al., ‘Blood-based biomarkers of Alzheimer’s disease and incident dementia in the community,’ Nature Medicine, vol. 31, no. 6, p. 2027–2035, 2025. [166] S. Palmqvist, H. E. Whitson, L. A. Allen et al., ‘Alzheimer’s Association Clinical Practice Guideline on the use of blood-based biomarkers in the diagnostic workup of suspected Alzheimer’s disease within specialized care settings,’ Alzheimer’s & Dementia, vol. 21, no. 7, p. e70535, 2025. [167] A. L. Benedet, M. Milà-Alomà, A. Vrillon, N. J. Ashton, T. A. Pascoal et al., ‘Differences between plasma and cerebrospinal fluid glial fibrillary acidic protein levels across the alzheimer disease continuum,’ JAMA Neurology, vol. 78, no. 12, p. 1471–1483, 2021. [168] N. Mattsson, U. Andreasson, H. Zetterberg, and K. Blennow, ‘Associ- ation of plasma neurofilament light with neurodegeneration in patients with alzheimer disease,’ JAMA Neurology, vol. 74, no. 5, p. 557–566, 2017. [169] N. J. Ashton, S. Janelidze, A. Al Khleifat, A. Leuzy, E. L. van der Ende et al., ‘A multicentre validation study of the diagnostic value of plasma neurofilament light,’ Nature Communications, vol. 12, no. 1, p. 3400, 2021. [170] M. Garg, S. Hejazi, S. Fu et al., ‘Characterizing the progression from mild cognitive impairment to dementia: A network analysis of longitudinal clinical visits,’ BMC Medical Informatics and Decision Making, vol. 24, p. 305, 2024. [Online]. Available: https://doi.org/10. 1186/s12911-024-02711-z [171] G. J. Christie, T. Hamilton, B. D. Manor, N. A. S. Farb, F. Farzan, A. Sixsmith, J.-J. Temprado, and S. Moreno, ‘Do lifestyle activities protect against cognitive decline in aging? a review,’ Frontiers in Aging Neuroscience, vol. 9, p. 381, 2017. [Online]. Available: https://doi.org/10.3389/fnagi.2017.00381 [172] J. Alty, M. Farrow, and K. Lawler, ‘Exercise and dementia prevention,’ Practical Neurology, vol. 20, no. 3, p. 234–240, 2020. [Online]. Available: https://pn.bmj.com/content/20/3/234 [173] V. Alanko, C. Udeh-Momoh, M. Kivipelto, and A. Sandebring-Matton, ‘Mechanisms underlying non-pharmacological dementia prevention strategies: A translational perspective,’ The Journal of Prevention of Alzheimer’s Disease, vol. 9, no. 1, p. 3–11, 2021. [Online]. Available: https://doi.org/10.14283/jpad.2022.9 [174] L. J. Dominguez, N. Veronese, L. Vernuccio, G. Catanese, F. Inzerillo, G. Salemi, and M. Barbagallo, ‘Nutrition, physical activity, and other lifestyle factors in the prevention of cognitive decline and dementia,’ Nutrients, vol. 13, no. 11, p. 4080, 2021. [Online]. Available: https://doi.org/10.3390/nu13114080 [175] R. A. Nianogo, N. A. Churchill, and D. E. Barnes, ‘Hypothetical midlife lifestyle interventions and dementia risk in us older adults: A comparative effectiveness microsimulation study,’ Alzheimer’s & Dementia, vol. 20, p. e090580, 2024. [Online]. Available: https: //doi.org/10.1002/alz.090580 [176] T. Yamasaki, ‘Preventive strategies for cognitive decline and dementia: Benefits of aerobic physical activity, especially open-skill exercise,’ Brain Sciences, vol. 13, no. 3, p. 521, 2023. [Online]. Available: https://doi.org/10.3390/brainsci13030521 [177] P. Müllers, M. Taubert, and N. G. Müller, ‘Physical exercise as personalized medicine for dementia prevention?’ Frontiers in Physiology, vol. 10, p. 451073, 2019. [Online]. Available: https://doi.org/10.3389/fphys.2019.00672 [178] Z. Biskupiak, V. V. Ha, A. Rohaj, and G. Bulaj, ‘Digital therapeutics for improving effectiveness of pharmaceutical drugs and biological products: Preclinical and clinical studies supporting development of drug + digital combination therapies for chronic diseases,’ Journal of Clinical Medicine, vol. 13, no. 2, p. 403, 2023. [Online]. Available: https://doi.org/10.3390/jcm13020403 [179] E. Moll Van Charante, M. P. Hoevenaar-Blom, M. Song, S. Andrieu, N. Coley, J. Georges, W. A. Van Gool, R. Handels, Y. Wang, A. Wimo, C. B. Brayne, W. Wang, E. Richard, and PRODEMOS study group, ‘Prevention of dementia using mobile phone applications (prodemos): A multinational randomised controlled effectiveness-implementation trial,’ European Heart Journal, vol. 45, no. Supplement_1, p. ehae666.3555, October 2024. [Online]. Available: https://doi.org/10.1093/eurheartj/ ehae666.3555 [180] X. Hu, K. Karako, P. Song, W. Tang, and Y. Xia, ‘Guardians of memory: The urgency of early dementia screening in an aging society,’ Intractable & Rare Diseases Research, vol. 13, no. 3, p. 133, 2024. [Online]. Available: https://doi.org/10.5582/irdr.2024.01026 [181] C.-N. Kim and S.-H. Park, ‘Super-aged society and dementia: Community-based integrated management strategies learned from the kobe project,’ Journal of Korean Society of Neurocognitive Rehabilitation, vol. 16, no. 2, p. 11–19, 2024. [Online]. Available: https://doi.org/10.29144/KSCTE.2024.16.2.11 [182] L. van Rosmalen, C. C. Brück, F. J. Wolters, R. Handels, and I. M. C. M. de Kok, ‘Comparing the short-term and long-term impact of primary and secondary prevention strategies on population-level dementia burden: A microsimulation modelling study,’ Alzheimer’s & Dementia, vol. 20, p. e091332, 2024. [Online]. Available: https://doi.org/10.1002/alz.091332 [183] C. H. van Dyck, R. Sperling, K. Johnson et al., ‘Long-term safety and efficacy of lecanemab in early Alzheimer’s disease: Results from the Clarity AD open-label extension study,’ Alzheimer’s & Dementia, vol. 21, no. 12, p. e70905, 2025. [184] J. A. Zimmer, J. R. Sims, C. D. Evans et al., ‘Donanemab in early symp- tomatic Alzheimer’s disease: results from the TRAILBLAZER-ALZ 2 long-term extension,’ The Journal of Prevention of Alzheimer’s Disease, vol. 13, no. 2, p. 100446, 2026. [185] L. S. Schneider, R. E. Kennedy, and G. R. Cutter, ‘Caution in interpret- ing disease-modification claims with lecanemab: selective reporting and causal inference,’ Alzheimer’s & Dementia, vol. 22, no. 5, p. e71486, 2026. [186] J. Hazan, K. Y. Liu, and R. Howard, ‘Long-term extension data do not robustly support clinical disease-course modification with donanemab,’ The Journal of Prevention of Alzheimer’s Disease, vol. 13, no. 4, p. 100511, 2026. [187] S. Walsh, L. Wallace, I. Kuhn, O. Mytton, L. Lafortune, W. Wills, N. Mukadam, and C. Brayne, ‘Population-level interventions for the primary prevention of dementia: A complex evidence review,’ EClinicalMedicine, vol. 70, p. 102538, 2024. [Online]. Available: https://doi.org/10.1016/j.eclinm.2024.102538 22VOLUME 11, 2023 [188] M. Asif, D. Srivastava, A. Gupta, and U. S. Tiwary, ‘Inter subject emotion recognition using spatio-temporal features from EEG signal,’ arXiv preprint arXiv:2305.19379, 2023. [Online]. Available: https: //doi.org/10.48550/arXiv.2305.19379 [189] B. Zhang, Z. Fei, W. Zhou, and M. Fei, ‘A systematic review of artificial intelligence techniques for early detection of mild cognitive impairment in the elderly,’ in Proceedings of the 4th International Conference on Artificial Intelligence and Computer Engineering, ser. ICAICE ’23. New York, NY, USA: Association for Computing Machinery, 2024, p. 1096–1101. [Online]. Available: https://doi.org/10. 1145/3652628.3652808 [190] G. S. Collins, K. G. M. Moons, P. Dhiman, R. D. Riley, A. L. Beam, B. Van Calster, M. Ghassemi, X. Liu, J. B. Reitsma, M. van Smeden et al., ‘TRIPOD+AI statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods,’ BMJ, vol. 385, p. e078378, 2024. [191] J. Mongan, L. Moy, and C. E. Kahn, ‘Checklist for artificial intelligence in medical imaging (CLAIM): A guide for authors and reviewers,’ Radi- ology: Artificial Intelligence, vol. 2, no. 2, p. e200029, 2020. [192] P. Zanini, M. Congedo, C. Jutten, S. Said, and Y. Berthoumieu, ‘Transfer learning: A Riemannian geometry framework with applications to brain– computer interfaces,’ IEEE Transactions on Biomedical Engineering, vol. 65, no. 5, p. 1107–1116, 2018. [193] N. Ali, M. Asif, A. Kaushal, U. Singh, and U. S. Tiwary, ‘Optimizing emotion recognition in EEG data: A genetic algorithm approach with XAI insights,’ in 2024 15th International Conference on Computing Communication and Networking Technologies (ICCCNT). IEEE, 2024, p. 1–6. [Online]. Available: https://doi.org/10.1109/ICCCNT61001. 2024.10725514 [194] M. Ben Gara Ali and A. Smiti, ‘Privacy-preserving multimodal fusion for Alzheimer’s staging: a federated vision transformer framework with explainable AI,’ Computerized Medical Imaging and Graphics, vol. 129, p. 102730, 2026. [195] S.-J. Eun, E. J. Kim, and J. Y. Kim, ‘Development and evaluation of an artificial intelligence–based cognitive exercise game: A pilot study,’ Journal of Environmental and Public Health, vol. 2022, p. 4403976, 2022, 15 pages. [Online]. Available: https://doi.org/10.1155/2022/4403976 [196] P. Choudhary, M. Asif, and D. Srivastava, ‘Computational neuroscience: Simulating the systems biology of synaptic plasticity,’ in Synaptic Plasticity in Neurodegenerative Disorders. CRC Press, 2024, p. 226– 255. [Online]. Available: https://doi.org/10.1201/9781003464648-13 [197] M. Asif, D. Srivastava, and P. Choudhary, ‘Tools for improving synaptic plasticity from computational neuroscience,’ in Synaptic Plasticity in Neurodegenerative Disorders. CRC Press, 2024, p. 191–225. [Online]. Available: https://doi.org/10.1201/9781003464648-11 [198] M. Asif, S. Mishra, J. Kannath, T. Jayadevan, D. Singh, G. Goyal, A. Bhuyar, and U. S. Tiwary, ‘Cross cultural comparison of emotional functional networks,’ in Intelligent Human Computer Interaction (IHCI 2023), ser. Lecture Notes in Computer Science. Springer, 2023, p. 106– 114. [Online]. Available: https://doi.org/10.1007/978-3-031-53827-8_11 [199] M. El-Sayed, ‘A review of machine learning techniques for early detection of alzheimer’s disease,’ Metaheuristic Optimization Review, p. 22–36, 2024. [Online]. Available: https://doi.org/10.54216/MOR. 010203 MOHAMMAD ASIF is a Postdoctoral Fellow at the Indian Institute of Technology (IIT) Bombay, Mumbai, India, and a Senior Consultant at T- Systems ICT India Pvt. Ltd., Pune, India. He received the Ph.D. degree in artificial intelligence with a research focus on affective computing and EEG-based emotion recognition. His research integrates deep learning, fuzzy systems, and ex- plainable AI to model emotions in the valence- arousal-dominance (VAD) space. His work in- cludes the development of novel deep fuzzy frameworks, multimodal emotion-recognition models, and naturalistic EEG datasets for advancing human-centered AI. His current research interests include affective comput- ing, brain-computer interfaces, explainable and trustworthy AI, multimodal learning, agentic AI, and AI applications in healthcare and cognitive neuroscience. AZIZUDDIN KHAN is a Professor of psychology with the Department of Humanities and Social Sciences, Indian Institute of Technology Bombay, Mumbai, India, where he leads the Psychophys- iology Laboratory. His work integrates behav- ioral and neurophysiological methods, including electroencephalography (EEG) and event-related potentials (ERP), to investigate human cognition, learning, and brain function. His research inter- ests include cognitive psychology, cognitive neu- roscience, psychophysiology, developmental neuropsychology, working and prospective memory, developmental dyslexia, and cognitive ergonomics, with applications in education, healthcare, and cognitive disorders. MOHD AZAM is a Senior AI Solution Architect at T-Systems ICT India Pvt. Ltd., Pune, India, with more than 22 years of experience in infor- mation technology, including more than a decade specializing in artificial intelligence and machine learning. His work spans generative AI, large lan- guage models (LLMs), retrieval-augmented gen- eration (RAG), agentic AI, multi-agent systems, and enterprise AI architecture, with applications across telecommunications, banking, government, and smart-city domains. His research interests include agentic AI, enter- prise AI platforms, responsible AI, AI governance, explainable AI (XAI), trustworthy and scalable LLM-based systems, and autonomous intelligent agents. He is particularly interested in bridging AI research and industrial practice by developing secure, interpretable, and production-ready AI solutions that enable enterprise-scale digital transformation. VOLUME 11, 202323 ANURAG RAJKUMAR BOMBARDE is an En- terprise AI Architect and applied AI researcher with 19 years in software engineering, including more than 8 years in AI/ML and data science (2018–present) and more than 2.5 years building production generative and agentic AI systems. He received the T-Systems Trendmatcher Annual MD Award 2025 for outstanding contribution to AI, and has served as sole architect of enterprise AI platforms delivering over EUR 2 million in annual savings and 80–90% cycle-time reductions across cloud migration, finance, sales, QA automation, analytics, legal, HR, and document intelli- gence. His research interests center on verifiable and auditable AI, includ- ing cryptographic vector-commitment schemes (Verkle-tree constructions) for tamper-evident provenance, hypergraph representations of multi-party evidence and lineage, and the grounding of generated claims in standard- ized clinical vocabularies (SNOMED-CT, RxNorm) and interoperability standards (HL7 FHIR) for healthcare informatics. A parallel interest lies in human-AI cognition: distributed cognition across human-agent teams, and how trust calibration, provenance visibility, and explanation design shape whether domain experts appropriately accept or override machine output, extending to the cognitive-load effects of agentic automation on enterprise practitioners and the interface conditions under which auditability becomes usable rather than merely available. This work is complemented by hands- on depth in agentic architecture, including multi-agent orchestration (Lang- Graph, CrewAI, LangChain, LlamaIndex), subagent topologies, Agent Skills, loop and context engineering, and the Model Context Protocol (MCP) and agent-to-agent (A2A) protocols, together with knowledge- graph retrieval-augmented generation (Neo4j, FalkorDB, property-graph retrieval), organizational knowledge systems, rigorous LLM evaluation, AI security, ethics, and guardrail design. 24VOLUME 11, 2023