Vocal features based Parkinson's detection: An ensemble learning approach.

Area of Science:

• Neurology
• Artificial Intelligence

Background:

• Parkinson's disease (PD) is a progressive neurodegenerative disorder impacting the central nervous system.
• Global PD cases exceeded 8.5 million in 2019, highlighting the critical need for early detection and intervention.

Purpose of the Study:

• To identify an optimal machine learning technique for the early prediction of Parkinson's disease.
• To evaluate the efficacy of various machine learning algorithms using vocal features for PD detection.

Main Methods:

• Comparison of machine learning algorithms including Random Forest, K Nearest Neighbor, Naïve Bayes, Gradient Boosting, and XGBoost.
• Evaluation based on performance metrics such as recall, log loss, and overfitting resistance.
• Utilizing vocal features from large-scale datasets for PD prediction.

Main Results:

• Gradient Boosting demonstrated superior performance compared to other algorithms.
• The Gradient Boosting model achieved high recall, low log loss, and resistance to overfitting.
• Vocal features were identified as significant indicators for early-stage Parkinson's detection.

Conclusions:

• Machine learning, particularly Gradient Boosting, shows significant promise for early Parkinson's disease detection.
• Vocal biomarkers combined with machine learning can enhance diagnostic capabilities.
• This research facilitates medical centers by offering an optimized machine learning technique for PD diagnosis and decision-making.