Feature Aggregation for Alzheimer's Disease Diagnosis Using FDG-PET Images: The Potential of Graph-Based Methods
23rd IEEE International Symposium on Biomedical Imaging, ISBI 2026, London, İngiltere, 8 - 11 Nisan 2026, cilt.2026-April, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Cilt numarası: 2026-April
- Doi Numarası: 10.1109/isbi61048.2026.11515914
- Basıldığı Şehir: London
- Basıldığı Ülke: İngiltere
- Anahtar Kelimeler: Alzheimer's disease, FDG-PET, Feature ranking, Graph-based consensus, Rank Aggregation
- Galatasaray Üniversitesi Adresli: Hayır
Özet
Feature ranking for Alzheimer's disease (AD) classification often suffers from instability and inconsistent performance due to variations across neuroimaging datasets and analytic approaches. This study evaluates individual feature-ranking methods, simple aggregation techniques, and graph-based consensus approaches for AD classification. Significant discrepancies across individual ranking methods reveal their distinct sensitivities to data characteristics, underscoring the need for robust integration strategies. Aggregation-based approaches effectively leverage complementary insights to enhance stability and predictive accuracy. Among them, graph-based consensus methods-particularly our proposed multi-cutoff graph representations with Sinkhorn aggregation-demonstrate a principled capacity to capture complex feature dependencies and yield interpretable, high-fidelity consensus rankings. By advancing reliable feature selection and improving model transparency, this work contributes to strengthening computational tools that support early disease detection and facilitate more informed clinical decisionmaking in neurodegenerative disorders.