Alzheimer classification improves when brain evidence guides MRI and PET analysis
Evidence Before Accuracy: A MRI-PET Fusion Network for Alzheimer Disease Classification with Causal Regional Validation
Computer Vision and Pattern Recognition
Summary
Identifying Alzheimer disease from brain scans is tricky because models can rely on misleading clues unrelated to the disease. The authors developed a method that combines two types of brain images (MRI and PET) from specific brain areas known to be affected by Alzheimer’s. Their method checks that the model uses relevant brain regions, not shortcuts. This approach achieves strong accuracy while confirming that the brain areas involved match known disease biology.
What this means in practice
- •For medical imaging teams: Develop Alzheimer diagnostic tools that explicitly validate the brain regions driving decisions, improving trustworthiness.
- •For medical device manufacturers: Incorporate cost-effective fusion of MRI and PET images for improved Alzheimer classification into imaging systems.$Commercial implications: Enables building commercial brain imaging products providing accurate and biologically informed Alzheimer diagnoses.
Tested on one dataset.
Authors
Saeid Firouzi Daghigh, Saeed Ayat
Abstract
Deep learning models for Alzheimer disease (AD) classification routinely report near-perfect discrimination, yet few are shown to rest on AD-relevant neurobiology rather than on dataset artifacts, subject-level leakage, or non-brain image content. We present a fusion network combining T1 MRI and FDG PET across axial, coronal, and sagittal planes, trained on ADNI consists of 554 paired subjects. The fusion model reaches AUC 0.962, accuracy 0.909, and F1 0.891, competitive with recent 3D CNN and multimodal transformer systems at substantially lower cost. We first quantify how much modality, plane and slice geometry matter. A validation-only search over slice centres and neighbour spacings moves AUC by 0.180 for MRI and 0.078 for PET, selecting narrow spacing for MRI and wide spacing for PET, with the chosen coronal centres falling on the hippocampal body and on the posterior cingulate respectively. The contribution, however, is the evidence layer built around that number. Shortcut controls collapse the model to AUC 0.622 (silhouette), 0.608 (exterior), and 0.500 (blank), and a label-permutation null yields 0.456. Forward region-of-interest (ROI) ablation shows that masking medial temporal cortex in MRI and the posterior default-mode network (DMN) in PET produces the largest shift in the AD logit, while area-matched controls remain indistinguishable from that null. Reverse ROI ablation shows that the medial temporal lobe alone retains 89.2% of above-chance discrimination in MRI and the posterior DMN alone retains 79.0% in PET. A quantitative comparison of attribution methods shows occlusion sensitivity reaching 3.5-5.0* enrichment inside a priori AD regions against 0.10-0.43* in controls. Ablation and attribution independently establish a biologically correct double dissociation: hippocampal evidence is carried by MRI, posterior cingulate evidence by PET.