Veeky Baths

Dementia is a progressive neurodegenerative disorder with multiple etiologies, including Alzheimer's disease, Parkinson's disease, frontotemporal dementia, and vascular dementia. Its clinical and biological heterogeneity makes diagnosis and subtype differentiation highly challenging. Graph Neural Networks (GNNs) have recently shown strong potential in modeling brain connectivity, but their limited robustness, data scarcity, and lack of interpretability constrain clinical adoption. Explainable Graph Neural Networks (XGNNs) have emerged to address these barriers by combining graph-based learning with interpretability, enabling the identification of disease-relevant biomarkers, analysis of brain network disruptions, and provision of transparent insights for clinicians. This paper presents the first comprehensive review dedicated to XGNNs in dementia research. We examine their applications across Alzheimer's disease, Parkinson's disease, mild cognitive impairment, and multi-disease diagnosis. A taxonomy of explainability methods tailored for dementia-related tasks is introduced, alongside comparisons of existing models in clinical scenarios. We also highlight challenges such as limited generalizability, underexplored domains, and the integration of Large Language Models (LLMs) for early detection. By outlining both progress and open problems, this review aims to guide future work toward trustworthy, clinically meaningful, and scalable use of XGNNs in dementia research.

Marketing plays a crucial role in enhancing consumer awareness and affinity towards a product. Consumer research examines the motivations, trends, and patterns underlying purchasing behavior, thereby enabling institutions to refine their offerings and marketing strategies to improve their alignment with consumer decision-making and increase purchase probability. Consumer neuroscience aims to study the neurological factors influencing consumer preference, which cannot be determined using traditional methods like questionnaires and self-assessment. Machine learning techniques have proven effective in this domain, yet deep learning architectures that effectively fuse electroencephalography (EEG) signals with eye-tracking (ET) data remain largely unexplored. In this study, we propose a novel hybrid "Wide & Deep" framework for multimodal classification of consumerpreference using EEG and eye-tracking data from the Neuma dataset. The proposed architecture employs an InceptionTime network with early fusion to process concatenated multimodal input (25 channels), extracting multi-scale temporal featuresthrough parallel convolutions with varying kernel sizes (10, 20, 40). The learned 128-dimensional neural embeddings are combined with 44 time-domain statistical features (temporal mean and standard deviation across all channels) and passed to a CatBoost gradient boosting classifier, forming a neural-symbolic ensemble. A modality dropout strategy (30% probability) is introduced during training to prevent over-reliance on eye-tracking signals and force robust EEG representation learning. The framework employs a grokking-oriented training regime with extended epochs (100), high weight decay (0.02), and strict checkpointing based on validation loss. The model is evaluated using 5-fold stratified cross-validation to assess generalizability. The proposed framework achieves a mean accuracy of 77.51% (±0.49%) and a weighted F1-score of 74.60% (±0.21%) in classifying consumer preferences, representing a significant improvement over prior methods on this dataset. This study demonstrates the effectiveness of multi-scale convolutional architectures combined with gradient boosting ensemble methodsfor multimodal consumer neuroscience applications.

Background: Identifying early brain-based markers of cognitive decline is critical for preventive strategies in Alzheimer’s disease. Individuals with a familial risk may exhibit subtle functional brain changes years before clinical symptoms emerge. This exploratory study examined whether baseline functional brain network topology differentiates high-risk cognitively normal older adults who later progress to mild cognitive impairment (MCI) from those who remain cognitively stable. Methods: Baseline resting-state fMRI data were analyzed from 90 cognitively normal adults with a family history of Alzheimer’s (PREVENT-AD cohort), classified longitudinally as converters (MCI-C; n=45) or non-converters (MCI-NC; n=45). Whole-brain functional networks were analyzed across multiple thresholds; primary results are reported at 12% network density, with robustness verified at 16% density. Group differences were assessed using ANCOVA or Rank ANCOVA (controlling for age, sex, and education) at an uncorrected threshold (p < 0.05). Predictive utility was evaluated via a 100-repetition nested cross-validation machine-learning framework on a multimodal feature set combining functional network metrics, average cortical thickness, and plasma p-tau217, with covariates included within training folds. Results: At baseline, MCI-C participants were older, had fewer years of education, exhibited higher plasma p-tau217 levels, and showed trend-level lower MoCA scores. At 12% density, MCI-C showed increased average nodal strength (F=4.50, p=0.036, ηp ² =0.050) and reduced global efficiency (F=4.07, p=0.046, ηp ² =0.045). Increased betweenness centrality within the Default Mode Network (F=4.07, p=0.046, ηp ² =0.045) and trend-level increases in average clustering (F=3.10, p=0.081, ηp ² =0.035) were observed. Initial largest connected component (LCC) showed a trend-level decrease (F=3.84, p=0.053, ηp ² =0.043). At 16% density, MCI-C exhibited significantly reduced initial LCC (F=4.41, p=0.038, ηp ² =0.049) and increased nodal strength (F=4.29, p=0.041, ηp ² =0.048), with directionally consistent trend-level reductions in global efficiency (F=3.74, p=0.056, ηp ² =0.042). In machine learning, the k-nearest neighbors classifier showed the most stable performance (nested CV accuracy=59.6%; test F1-score=0.56). Feature stability analysis identified global efficiency (selected in 25.8% of iterations) and critical drop (19.4%) as the most consistent predictors. Conclusion: Baseline disruptions in functional network integration precede clinical conversion to MCI. The consistent selection of graph-theoretical metrics, particularly global efficiency and critical drop, as top predictors suggests that functional network reorganization provides unique information for classification before widespread cortical atrophy emerges.

Cognitive Assessment Games (CAGs) represent an emerging area of research. To date, most evidence regarding the efficacy of these games comes from studies c

Magnetic Resonance Imaging (MRI) plays a crucial role in detecting structural and functional abnormalities in the brains of epilepsy patients. Automating this process using advanced AI/ML techniques can reduce the burden on experts, be a training resource, and make diagnosis accessible and affordable to a larger population. However, models developed using imaging data have not achieved acceptable accuracy levels. One reason for this is the limited number of data points available and artifacts in the data capture mechanism of the available datasets. There is a need to develop a generalized model that can take as input diverse images with multiple conditions, which predicts epilepsy and identifies the impacted brain regions. In this work, we present the design of deep-learning models capable of predicting epilepsy using 3D brain Magnetic Resonance Imaging (MRI) scans. Five deep learning models were developed to classify 3D MRI scans to identify specific brain regions involved in Hippocampal Sclerosis, Dysembryoplastic Neuroepithelial Tumors, Cavernomas, Focal Cortical Dysplasia, and other epileptic-related conditions. We find that our trained models predict epilepsy with over 91% accuracy on FLAIR images and nearly 100% accuracy on T1-weighted (T1-w) images. Our results show that deep learning architectures in 3D are highly effective in detecting epilepsy and identifying relevant pathological regions. These findings highlight not only the diagnostic accuracy of the models but also their potential utility in aiding clinical decision-making.

Introduction: Early‐Onset Dementia (EOD), including Frontotemporal Dementia (FTD), behavioral variant FTD (bvFTD), and Early‐Onset Alzheimer’s Disease (EOAD), presents significant diagnostic and therapeutic challenges due to heterogeneous clinical features and rapid progression. EOD involves distinct patterns of brain network disruption, measurable through graph‐theoretical analysis. Methods: We reviewed 23 studies applying graph theory to electroencephalography (EEG), functional MRI (fMRI), diffusion tensor imaging (DTI), and fluorodeoxyglucose positron emission tomography (FDG‐PET) in EOD populations. Metrics included global and local efficiency, small‐world properties, modularity, hub connectivity, and rich‐club organization. Results: EOD demonstrates widespread topological disruption, including reduced global and local efficiency, hub vulnerability, and modular fragmentation. Subtype‐specific patterns include compensatory efficiency increases and parietal‐to‐frontal hub shifts in FTD; disruption of hub connectivity and modular integrity within the salience network in bvFTD; and pronounced deterioration of the default‐mode network in EOAD. Small‐world properties are generally preserved in early stages, reflecting initial compensatory reorganization that precedes system‐wide collapse. Conclusion: Graph‐theoretical analysis reveals characteristic topological disruptions in EOD. However, most studies rely on static measures, limiting insight into dynamic network vulnerability. Incorporating network resilience based computational models, longitudinal designs, and standardized analytical pipelines could clarify network failure mechanisms, uncover latent fragilities, and guide targeted interventions.

Dementia is a progressive neurodegenerative disorder with multiple etiologies, including Alzheimer's disease, Parkinson's disease, frontotemporal dementia, and vascular dementia. Its clinical and biological heterogeneity makes diagnosis and subtype differentiation highly challenging. Graph Neural Networks (GNNs) have recently shown strong potential in modeling brain connectivity, but their limited robustness, data scarcity, and lack of interpretability constrain clinical adoption. Explainable Graph Neural Networks (XGNNs) have emerged to address these barriers by combining graph-based learning with interpretability, enabling the identification of disease-relevant biomarkers, analysis of brain network disruptions, and provision of transparent insights for clinicians. This paper presents the first comprehensive review dedicated to XGNNs in dementia research. We examine their applications across Alzheimer's disease, Parkinson's disease, mild cognitive impairment, and multi-disease diagnosis. A taxonomy of explainability methods tailored for dementia-related tasks is introduced, alongside comparisons of existing models in clinical scenarios. We also highlight challenges such as limited generalizability, underexplored domains, and the integration of Large Language Models (LLMs) for early detection. By outlining both progress and open problems, this review aims to guide future work toward trustworthy, clinically meaningful, and scalable use of XGNNs in dementia research.

Despite a strong desire to quit, individuals with long-term substance use disorder (SUD) often struggle to resist drug use, even when aware of its harmful consequences. This disconnect between knowledge and compulsive behavior reflects a fundamental cognitive-behavioral conflict in addiction. Neurobiologically, differential cue-induced activity within striatal subregions, along with dopamine-mediated connectivity from the ventral to the dorsal striatum, contributes to compulsive drug-seeking. However, the functional mechanism linking these findings to behavioral conflict remains unclear. Another hallmark of addiction is temporal discounting: individuals with drug dependence exhibit steeper discount rates than non-users. Assuming the ventral-dorsal striatal organization reflects a gradient from cognitive to motor representations, addiction can be modeled within a hierarchical reinforcement learning (HRL) framework. However, integrating discounting into biologically grounded HRL remains an open challenge. In this work, we build on a model showing how action choices reinforced with drug rewards become insensitive to the negative consequences that follow. We address the integration of discounting by ensuring natural reward values converge across all levels in the HRL hierarchy, while drug rewards diverge due to their dopaminergic effects. Our results show that high discounting amplifies drug-seeking across the hierarchy, linking faster discounting with increased addiction severity and impulsivity. We demonstrate alignment with empirical findings on temporal discounting and propose testable predictions, establishing addiction as a disorder of hierarchical decision-making.

Cognition-assessing games (CAGs) are an evolving area of research, with collaboration between game developers and neuroscience researchers playing a critical role in their development. Game developers bring valuable expertise in game design and development, while neuroscience researchers contribute their knowledge of cognitive assessment. However, CAGs differ from both mainstream entertainment games (game developers’ expertise) and the traditional neuropsychological assessments (neuroscience researchers’ expertise). Given this context, it is essential to address the ideological challenges and conflicts that may arise between game developers and neuroscience researchers as they collaborate on CAG development. In this viewpoint, we present the perspectives of both game developers and neuroscience researchers on ten key themes relevant to CAGs: game intent, game tutorial, game art style and user interface, learning effects, ecological validity, game scoring, game realism, game testing, commercialization, and game platform. These themes are identified based on our over four years of collaborative work and are further supported by existing literature on CAGs. Our viewpoint is grounded in practical examples, offering concrete insights to facilitate impactful development of CAGs. The ten themes provide a structured framework for decision-making in CAGs emphasizing the importance of a systematic approach to CAG development. Finally, we hope that the findings from our viewpoint will assist game developers and neuroscience researchers in creating successful, translatable CAGs.