Uncertainty-aware personalized estimation of Parkinson's disease severity from longitudinal speech.

Parkinson's disease (PD) is a progressive neurological disorder characterized by motor impairments whose severity is commonly assessed using the Unified Parkinson's Disease Rating Scale (UPDRS). Although clinically established, UPDRS assessment is inherently subjective, requiring in-person evaluation by trained specialists, limiting its suitability for frequent monitoring. Speech production is affected early in PD and provides a non-invasive modality for remote symptom assessment. In this study, an uncertainty-aware personalized framework is proposed for estimating PD severity from speech signals. The approach integrates longitudinal temporal modeling of longitudinal speech recordings with patient-specific representations and a probabilistic latent disease state. Continuous motor UPDRS scores are jointly estimated with data-driven ordinal disease severity stages, enabling both fine-grained regression and auxiliary ordinal prediction. Predictive uncertainty is explicitly quantified to characterize predictive variability within the proposed framework. The method is evaluated on a longitudinal speech dataset using a strict patient-wise split, ensuring that all test subjects are unseen during training. On the held-out test set, the proposed model achieves promising predictive accuracy (mean absolute error 0.56 UPDRS points, root mean squared error 0.74, and coefficient of determination R2 = 0.99) for motor UPDRS estimation. Ordinal severity classification attained an accuracy of 0.92 across three stages. Comparative experiments against classical machine learning methods and global temporal baselines demonstrate consistent performance improvements. These results demonstrate the potential of personalized, uncertainty-aware speech modeling for longitudinal PD severity estimation.