C-Atlas: A Brain Mapping based on FDG-PET Images for Alzheimer's Disease Diagnosis

Alzheimer's Disease (AD) poses many challenges to healthcare systems for elderly people in the world, for which an accurate diagnosis of AD is therefore a crucial task. Positron Emission Tomography (PET) images and machine learning can be used to support this purpose. Brain atlases are common tools used to extract features from PET images, and choosing a proper brain map is extremely important. In this work, we propose a novel method to learn the very first PET-driven brain mapping based on Support Vector Machine (SVM) coefficients and Simple Linear Iterative Clustering (SLIC) algorithm, which is addressed as C-Atlas. When compared with predefined atlases on classification task, C-atlas experimentally proved to be more adapted for AD diagnosis both in terms of feature extraction and selection. The results suggest that it is possible to build an adapted brain mapping for a specific disease, and that the proposed method can also be used as an effective mean to support the analysis and comparison of brain diseases at multiple scales.