Voice biomarkers as prognostic indicators for Parkinson's disease using machine learning techniques

Area of Science:

• Neurology
• Biomedical Engineering
• Data Science

Background:

• Parkinson's disease is a progressive neurological disorder affecting millions globally, characterized by dopamine deficiency and motor symptoms like tremors and rigidity.
• Current diagnostic methods can be challenging for early detection, highlighting the need for novel, non-invasive approaches.
• Vocal impairment is a common symptom in Parkinson's patients, suggesting its potential as an early indicator.

Purpose of the Study:

• To investigate the efficacy of vocal measures for the early prediction of Parkinson's disease.
• To compare the performance of various machine learning models in classifying Parkinson's patients based on voice data.
• To evaluate feature selection techniques for optimizing diagnostic accuracy.

Main Methods:

• Utilized a dataset of 195 vocal recordings from 31 individuals (Parkinson's patients and healthy controls).
• Applied machine learning algorithms including Support Vector Machine (SVM), Random Forest (RF), Logistic Regression (LR), and Decision Tree (DT).
• Employed Synthetic Minority Over-Sampling Technique (SMOTE) for class imbalance and Principal Component Analysis (PCA) for feature selection.

Main Results:

• Random Forest (RF) demonstrated the highest performance, achieving 94% accuracy and 94% precision.
• Support Vector Machine (SVM) achieved 92% accuracy and 91% precision without feature selection.
• After PCA, SVM, RF, and Decision Tree (DT) achieved accuracies of 89%, 92%, and 87%, respectively, indicating the impact of feature selection.

Conclusions:

• Vocal features, when analyzed with advanced machine learning, offer a reliable method for early Parkinson's disease diagnosis.
• Machine learning models, especially Random Forest, show significant potential in differentiating between healthy individuals and Parkinson's patients using voice data.
• This research underscores the importance of voice analysis as a cost-effective and accessible tool for early Parkinson's detection, addressing current diagnostic challenges.