Cross-Attention Transformer for Early Parkinson’s Detection

Created using ChatSlide

This coursework examines the groundbreaking advancements in early Parkinson's Disease (PD) diagnosis through the development of a novel model, XAMT++. Spanning from November 2025 to April 2026, the work by Jungpil Shin and colleagues addresses limitations in unimodal diagnostics by introducing a robust, multi-modal approach. Key findings highlight an accuracy rate of 0.87 and an AUROC of 0.89, showcasing significant improvements in early detection. Despite challenges with small datasets,...