This research introduces novel, machine learning-based speech biomarkers for classifying Parkinson's disease (PD) severity across multiple stages, utilizing an Indian speech corpus. The study leverages multidomain acoustic features, which are then calibrated against clinical labels, offering a potentially more nuanced and objective approach to assessing PD progression through speech. The development of these ML-based biomarkers is significant as it moves beyond simple binary classifications of PD. By enabling multiclass severity assessment, this work could lead to more precise monitoring of disease progression and potentially inform treatment adjustments. The use of an Indian corpus also highlights the importance of diverse datasets in developing generalizable speech-based diagnostic tools.
Original publication date: 2026 Apr 14
Read Full Research Article →