xEEGNet: towards explainable AI in EEG dementia classification

Topic: Risk, prevention and care in population studies

Abstract Objective. This work presents xEEGNet, a novel, compact, and explainable neural network for electroencephalography (EEG) data analysis. It is fully interpretable and reduces overfitting through a major parameter reduction. Approach. As an applicative use case to develop our model, we focused on the classification of common dementia conditions, Alzheimer’s and frontotemporal dementia, versus controls. xEEGNet, however, is broadly applicable to other neurological conditions involving spectral alterations. We used ShallowNet, a simple and popular model in the EEGNet family, as a starting point. Its structure was analyzed and gradually modified to move from a ‘black box’ model to a more transparent one, without compromising performance. The learned kernels and weights were analyzed from a clinical standpoint to assess their medical significance. Model variants, including ShallowNet and the final xEEGNet, were evaluated using a robust nested-leave-n-subjects out cross-validation for unbiased performance estimates. Variability across data splits was explained using the embedded EEG representations, grouped by class and set, with pairwise separability to quantify group distinction. Overfitting was measured through training-validation loss correlation and training speed. Main results. xEEGNet uses only 168 parameters, 200 times fewer than ShallowNet, yet retains interpretability, resists overfitting, achieves comparable median performance (−1.5%), and reduces performance variability across splits. This variability is explained by the embedded EEG representations: higher accuracy correlates with greater separation between test-set controls and Alzheimer’s cases, without significant influence from the training data. Significance. The capability of xEEGNet to filter specific EEG bands, learns band specific topographies and use the right EEG spectral bands for disease classification demonstrates its interpretability. While big deep learning models are typically prioritized for performance, this study shows that smaller architectures like xEEGNet can be equally effective in pathology classification, using EEG data.

Key findings

• As an applicative use case to develop our model, we focused on the classification of common dementia conditions, Alzheimer’s and frontotemporal dementia, versus controls. xEEGNet, however, is broadly applicable to other neurological conditions involving spectral alterations.
• Main results. xEEGNet uses only 168 parameters, 200 times fewer than ShallowNet, yet retains interpretability, resists overfitting, achieves comparable median performance (−1.5%), and reduces performance variability across splits.
• This variability is explained by the embedded EEG representations: higher accuracy correlates with greater separation between test-set controls and Alzheimer’s cases, without significant influence from the training data.

Why it matters

This work presents xEEGNet, a compact and explainable neural network that classifies Alzheimer's and frontotemporal dementia versus controls from EEG data with far fewer parameters than a comparison model. It is a diagnostic classification tool and does not treat, cure, or prevent dementia.

New to this subject? Start with our plain-English guide: Disparities →

Study details

• Effect size1.5%
• Evidence strengthLow

Read the original

• DOI: 10.1088/1741-2552/adf6e6
• PubMed: 40752516
• Original source →
• Free full text: Journal of Neural Engineering →

Full text: free to read — OpenAlex lists this copy as free to read: the link above (checked 6 Feb 1020).

@article{zanola2025-40752516,
  author  = {Zanola, Andrea and Tshimanga, Louis Fabrice and Del Pup, Federico and Baiesi, Marco and Atzori, Manfredo},
  title   = {{xEEGNet: towards explainable AI in EEG dementia classification}},
  journal = {Journal of Neural Engineering},
  year    = {2025},
  doi     = {10.1088/1741-2552/adf6e6},
  pmid    = {40752516},
  url     = {https://doi.org/10.1088/1741-2552/adf6e6}
}

Studies this one cites

Referenced by this study and also in the alzhub library, so you can read them here.

• The Montreal Cognitive Assessment, MoCA: a brief screening tool for mild cognitive impairment Journal of the American Geriatrics Society · 2005-04-01 · usefulness score 60/100 · Getting a diagnosis
• “Mini-mental state” Journal of Psychiatric Research · 1975-11-01 · usefulness score 55/100 · Cognitive tests and assessment
• Alzheimer's Disease and Frontotemporal Dementia: A Robust Classification Method of EEG Signals and a Comparison of Validation Methods Diagnostics (Basel, Switzerland) · 2021-08-09 · usefulness score 50/100 · Digital health tools
• Evaluating EEG complexity and spectral signatures in Alzheimer’s disease and frontotemporal dementia: evidence for rostrocaudal asymmetry npj Aging · 2025-06-09 · usefulness score 39/100 · Lewy body dementia

Related studies in Disparities

• Use of explainable AI (xAI) in dementia detection and prognosis: a scoping review BMC Medical Informatics and Decision Making · 2026-01-16 · usefulness score 57/100
• Comparison of Methods for Algorithmic Classification of Dementia Status in the Health and Retirement Study Epidemiology (Cambridge, Mass.) · 2019-03-01 · usefulness score 48/100
• Improving Early Detection of Dementia: Extra Trees-Based Classification Model Using Inter-Relation-Based Features and K-Means Synthetic Minority Oversampling Technique Big Data and Cognitive Computing · 2025-05-30 · usefulness score 35/100
• Qualitative study exploring knowledge and attitudes towards dementia risk prediction, barriers to dementia services and service improvement recommendations with diverse populations in England BMJ Open · 2025-05-01 · usefulness score 42/100
• Effects of clinical stage, behavioral and psychological symptoms of dementia, and living arrangement on social distance towards people with dementia PLoS ONE · 2025-01-22 · usefulness score 38/100
• Attitudes towards euthanasia among physicians, nurses and the general public in Finland Public health · 2002-11-01 · usefulness score 32/100

Also tagged “classification”

835 studies in the library carry the topic tag “classification”; these are this study's neighbours in a fixed, alphabetical-by-hash tour of them, not the closest in content. All 835 tagged studies.

• Activated MAO-B in the brain of Alzheimer patients, demonstrated by [11C]-L-deprenyl using whole hemisphere autoradiography Neurochemistry international · 2010-11-12 · usefulness score 43/100 · Biomarkers
• A longitudinal analysis of T1-weighted MRI features associated with progression from mild cognitive impairment to Alzheimer’s disease BMC Medical Imaging · 2026-05-11 · usefulness score 48/100 · Mild cognitive impairment
• AI-driven magnetoencephalography biomarkers in dementia risk prediction: current evidence, challenges and future perspectives Frontiers in Dementia · 2026-04-07 · usefulness score 39/100 · Getting a diagnosis

Similar studies

• Leveraging Explainable AI for Dementia Classification: A Machine Learning Approach 2025 · Digital health tools
• A novel multi-modal machine learning based approach for automatic classification of EEG recordings in dementia 2020 · Digital health tools
• Standardizing Number of EEG Sensors for AI-Driven Dementia Detection 2025 · Digital health tools
• EEG-Based Dementia Classification Using CS-EMD Synchrony Features and Quantum Machine Learning 2025 · Digital health tools
• Application of machine learning in EEG-based dementia diagnosis: Classification and differential diagnosis 2025 · Digital health tools
• Investigating the Utility of Explainable Artificial Intelligence for Neuroimaging‐Based Dementia Diagnosis and Prognosis 2026 · Digital health tools

More on Disparities

• Newest in this subject Racial Differences and Temporal Trends in Parkinson’s and Alzheimer’s Disease-Related medRxiv · 2026-10-24 · Preprint — not yet peer-reviewed · usefulness score 16/100
• Newest in this subject Prevalence of dementia and mild cognitive impairment in Morocco: A population-based study Research Square · 2026-10-22 · Preprint — not yet peer-reviewed · usefulness score 16/100

See all studies in Disparities →