Multi-model deep learning for dementia detection: addressing data and model limitations

Topic: Risk, prevention and care in population studies

Among all evaluated models, the 3D-CNN and multimodal attention models achieved the highest performance, with accuracies of 84% and 86%, specificities of 83% and 86%, and sensitivities of 84% and 86%, respectively. The ViT and CapsNet models achieved 100% sensitivity for Alzheimer's disease (AD) but demonstrated low precision for AD (43%) and 0% for other classes, indicating class imbalance effects. All models showed reduced performance and bias toward certain classes. The findings highlight the limitations of current architectures in sMRI dementia classification, including suboptimal feature extraction and class-specific biases. While certain models, such as multimodal attention and 3D-CNN, performed better overall, precision and generalization remain challenges. Future work should focus on improved data representation through advanced computer vision methods and architectural modifications to enhance diagnostic accuracy and computational efficiency.

Key findings

• Among all evaluated models, the 3D-CNN and multimodal attention models achieved the highest performance, with accuracies of 84% and 86%, specificities of 83% and 86%, and sensitivities of 84% and 86%, respectively.
• The ViT and CapsNet models achieved 100% sensitivity for Alzheimer's disease (AD) but demonstrated low precision for AD (43%) and 0% for other classes, indicating class imbalance effects.
• All models showed reduced performance and bias toward certain classes.
• The findings highlight the limitations of current architectures in sMRI dementia classification, including suboptimal feature extraction and class-specific biases.
• While certain models, such as multimodal attention and 3D-CNN, performed better overall, precision and generalization remain challenges.

Why it matters

This study tested several deep learning models for classifying dementia from structural MRI scans, and the best models reached about 84-86% accuracy. However, all models showed reduced performance and bias toward certain classes, so these tools are not yet reliable enough for diagnosis and need better data representation and architecture.

New to this subject? Start with our plain-English guide: Getting a diagnosis →

Study details

• Evidence strengthLow

Read the original

• DOI: 10.3389/fnins.2025.1638022
• PubMed: 41378346
• Original source →
• Free full text: Frontiers in Neuroscience →

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

@article{bayahya2025-41378346,
  author  = {Bayahya, Areej Y. and Jammal, Fares and Banjar, Haneen Reda and Eassa, Fathy Elbouraey and Talabay, Omar and Alamri, Sultan Hassan},
  title   = {{Multi-model deep learning for dementia detection: addressing data and model limitations}},
  journal = {Frontiers in Neuroscience},
  year    = {2025},
  doi     = {10.3389/fnins.2025.1638022},
  pmid    = {41378346},
  url     = {https://doi.org/10.3389/fnins.2025.1638022}
}

Studies this one cites

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

• Mini-Mental State Examination (MMSE) for the detection of Alzheimer's disease and other dementias in people with mild cognitive impairment (MCI) The Cochrane database of systematic reviews · 2015-03-05 · usefulness score 72/100 · Getting a diagnosis
• Structural magnetic resonance imaging for the early diagnosis of dementia due to Alzheimer's disease in people with mild cognitive impairment The Cochrane database of systematic reviews · 2020-03-02 · usefulness score 71/100 · Mild cognitive impairment
• A deep learning model for early prediction of Alzheimer's disease dementia based on hippocampal magnetic resonance imaging data Alzheimer's & dementia : the journal of the Alzheimer's Association · 2019-06-11 · usefulness score 56/100 · Prevention and risk

Related studies in Getting a diagnosis

• A deep joint-learning proteomics model for diagnosis of six conditions associated with dementia Nature medicine · 2026-03-31 · usefulness score 62/100
• Multi-modal data analysis for early detection of alzheimer’s disease and related dementias The Journal of Prevention of Alzheimer s Disease · 2025-12-01 · usefulness score 40/100
• Muti-band Morlet mutual information functional connectivity for classifying Alzheimer's disease and frontotemporal dementia with a deep learning technique Computers in Biology and Medicine · 2025-09-05 · usefulness score 38/100
• Mayo Normative Studies: Regression-based normative data for remote self-administration of the Stricker Learning Span, Symbols Test, and Mayo Test Drive Screening Battery Composite and validation in individuals with mild cognitive impairment and dementia The Clinical Neuropsychologist · 2025-03-11 · usefulness score 50/100
• A Machine Learning-Based Framework for differential screening of Neurodegenerative Dementias using routine clinical data Research Square · 2026-10-07 · usefulness score 16/100
• Investigating Brain Lobe Biomarkers to Enhance Dementia Detection Using EEG Data Cognitive Computation · 2025-04-01 · usefulness score 42/100

Similar studies

• Machine learning algorithms and traditional statistical models for detection of dementia: a population-based study 2025 · Digital health tools
• Enhanced Multi-Model Machine Learning-Based Dementia Detection Using a Data Enrichment Framework: Leveraging the Blessing of Dimensionality 2025 · Digital health tools
• Neuroimaging and Machine Learning for Dementia Diagnosis: Recent Advancements and Future Prospects 2019 · Digital health tools
• Leveraging explainable artificial intelligence with ensemble of deep learning model for dementia prediction to enhance clinical decision support systems 2025 · Digital health tools
• Machine-learning techniques for building a diagnostic model for very mild dementia 2010 · Digital health tools
• Investigating dementia patterns by machine learning and graph-based marker analysis for multiclass dementia diagnosis 2026 · Getting a diagnosis

More on Getting a diagnosis

• A review that covers this ground Montreal Cognitive Assessment (MoCA) Scale: Strengths, Limitations, and Implication for Clinical Practice Iranian Journal of Psychiatry · 2025-12-20 · usefulness score 32/100
• Newest in this subject The Contribution of Routinely Collected Clinical Data to Dementia Identification in Primary Care: A Cross-Sectional Study Research Square · 2026-10-28 · Preprint — not yet peer-reviewed · usefulness score 16/100
• Newest in this subject Beyond “Denial”: Understanding and Responding to Help-Refusal in Dementia Preprints.org · 2026-10-26 · Preprint — not yet peer-reviewed · usefulness score 16/100

See all studies in Getting a diagnosis →