An Explainable Deep Learning Approach for Frontotemporal Dementia Diagnostic Classification from Structural MRI

• Caio dos S. Felipe UFCSPA
• Dieine E. B. Schiavon UFCSPA
• Thatiane A. P. Alva UFCSPA
• Carla D. L. Becker UFCSPA

DOI: https://doi.org/10.5753/eniac.2026.30003

Resumo

This paper presents an explainable deep learning pipeline for fourclass diagnostic classification from T1-weighted structural MRI, discriminating three frontotemporal dementia subtypes from cognitively normal controls. A ConvNeXt-Tiny backbone trained on the Neuroimaging Initiative for Frontotemporal Dementia dataset processes 16 sagittal slices and achieves 84.2% accuracy and macro AUC of 0.942 on a held-out patient cohort. A learned Slice Attention Pooling module exposes per-slice aggregation weights, enabling positional inspection of volume-level decisions. EigenCAM and slice attention profiles yield anatomically consistent explanations for BV, PNFA, and CON, with SV discrimination encoded in the global embedding space.

Referências

Antonioni, A., Raho, E. M., Lopriore, P., Pace, A., Spina, E., et al. (2023). Frontotemporal dementia, where do we stand? a narrative review. International Journal of Molecular Sciences, 24(14):11732.

Grossman, M., Seeley, W. W., Boxer, A. L., et al. (2023). Frontotemporal lobar degeneration. Nature Reviews Disease Primers, 9(40).

Hu, J., Qing, Z., Liu, R., Zhang, X., Lv, P., Wang, M., et al. (2021). Deep learning-based classification and voxel-based visualization of frontotemporal dementia and alzheimer’s disease. Frontiers in Neuroscience, 14:626154.

Liu, Z., Mao, H., Wu, C.-Y., Feichtenhofer, C., Darrell, T., and Xie, S. (2022). A ConvNet for the 2020s. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 11976–11986.

Ma, D., Stocks, J., Rosen, H., Kantarci, K., Lockhart, S. N., Bateman, J. R., et al. (2024). Differential diagnosis of frontotemporal dementia subtypes with explainable deep learning on structural MRI. Frontiers in Neuroscience, 18:1331677.

Martin, S. A., Zhao, A., Qu, J., Imms, P., Irimia, A., Barkhof, F., and Cole, J. H. (2026). Investigating the utility of explainable artificial intelligence for neuroimaging-based dementia diagnosis and prognosis. Human Brain Mapping, 47(2):e70456.

Metz, A., Zeighami, Y., Ducharme, S., Villeneuve, S., and Dadar, M. (2025). Frontotemporal dementia subtyping using machine learning, multivariate statistics and neuroimaging. Brain Communications, 7(1):fcaf065.

Muhammad, M. B. and Yeasin, M. (2021). Eigen-CAM: Visual explanations for deep convolutional neural networks. SN Computer Science, 2(1):47.

National Institute on Aging (2021). Frontotemporal disorders: Information for patients, families, and caregivers.

NIFD (2010). NIFD: Neuroimaging in frontotemporal dementia database. Laboratory of Neuro Imaging (LONI), USC. [link] (accessed on 08 Jun 2026).

Singh, D., Brima, Y., Levin, F., Becker, M., Hiller, B., Hermann, A., et al. (2025). An unsupervised XAI framework for dementia detection with context enrichment. Scientific Reports, 15(1):39554.

Termine, A., Fabrizio, C., Caltagirone, C., and Petrosini, L. (2022). A reproducible deep-learning-based computer-aided diagnosis tool for frontotemporal dementia using MONAI and clinica frameworks. Life, 12(7):947.

Viswan, V., Shaffi, N., Mahmud, M., Subramanian, K., and Hajamohideen, F. (2024). Explainable artificial intelligence in alzheimer’s disease classification: A systematic review. Cognitive Computation, 16(1):1–44.

Whitwell, J. L. (2019). Ftd spectrum: Neuroimaging across the FTD spectrum. In Becker, J. T. and Cohen, A. D., editors, Brain Imaging, volume 165 of Progress in Molecular Biology and Translational Science, pages 187–223. Elsevier.