First Benchmark for Early-Stage Parkinson's Detection from Speech

A pioneering benchmark for identifying early-stage Parkinson's disease (EarlyPD) through speech has been established by researchers. This benchmark aims to resolve the inconsistencies across studies caused by variations in datasets, languages, tasks, evaluation methods, and definitions. It employs a speaker-independent approach to ensure fair and reproducible evaluations across accessible datasets. Covering three prevalent speech tasks, it assesses methods under diverse training-resource conditions. Detailed evaluations categorized by dataset, aggregation level, gender, and disease stage enable precise comparisons and support clinical implementation. The findings offer a replicable reference and practical insights, promoting the use of this publicly accessible benchmark to enhance effective and clinically relevant EarlyPD detection through speech.