Does Brain Network Construction Choice Matter? an Empirical Study of Individual Networks from Static FDG-PET for Alzheimer's Diagnosis

This paper presents a comparative analysis of multiple methods for constructing individual graphs from static FDG-PET images, with a focus on Alzheimer's Disease (AD) diagnosis. We evaluate five methods, categorized into mean-based and probability density function (PDF)-based graphs, and compare them using two criteria: structural similarity between individual and group-level graphs, and their effectiveness as features in AD classification tasks. Our findings demonstrate the superior performance of PDF-based methods in capturing individual variability and improving classification accuracy, while also identifying strengths and limitations of methods. This study underscores the importance of selecting appropriate graph construction techniques and offers valuable insights for enhancing AD diagnosis through graph-based analysis.