Uncertainty Quantification Is Indispensable for Reliable Connectome-Based Graph Learning: A Narrative Review and Case Study
Graph neural networks are used for brain imaging analysis but often produce overconfident predictions that don't match reality. This paper reviews uncertainty quantification (UQ) methods like Bayesian approaches and ensemble methods, using a case study where a GAT model showed high accuracy but severe overconfidence. It concludes that rigorous UQ and calibration are essential for trustworthy clinical use of graph biomarkers.