An analysis of vocal features for diagnosing Parkinson’s disease

Voice signals are increasingly studied as a potential biomarker for the diagnosis of the Parkinson’s disease (PD).
This study evaluated 195 voice signals from 31 patients using different machine learning tools. In an attempt to differentiate between people with PD and healthy individuals based on voice recordings, the authors found that multi-layer perceptron (MLP; overall accuracy= 98.31%; overall recall= 98%; overall precision= 100%) and support vector machine (SVM; overall accuracy= 95%; overall recall= 96%; overall precision= 98%) models offered the best results for the diagnosis of PD.
According to the authors, MLP and SVM machine learning tools performs effectively and can be applied in clinical practice for the diagnosis of PD.