Detecting Parkinson's Disease Using Voice Recordings from Mobile Devices

This study examines how one may detect Parkin-son's disease (PD) by analyzing voice recordings made with a mobile phone. The key objectives include creating a model that ensures fair and unbiased predictions while maintaining interpretability that is consistent with the existing literature on PD. We introduce a novel group-wise scaling method to address typical age and biological sex biases in the data sets, demonstrating an average 14% improvement as compared to conventional scaling. We also show the importance of evaluating the developed model on unseen individuals to achieve reliable classification results. The developed model offers an accuracy of 80.3% with a recall of 89.4% for unseen individuals, surpassing current state-of-the-art approaches. Furthermore, we offer insights into the decision-making of the model using SHAP values. Our analysis reveals that a low standard deviation in the slope of the fundamental frequency (indicative of monotone voice quality) increases the likelihood of classifying a PD voice. Additionally, longer unvoiced segments, increased loudness standard deviation, and jitter correlate with the presence of PD. The significance of our proposed model lies in its generalizability and reliability for early PD detection, potentially decelerating disease progression, reducing healthcare costs, and improving the quality of life for patients.