Automatic and Early Detection of Parkinson's Disease by Analyzing Acoustic Signals Using Classification Algorithms...

Area of Science:

• Biomedical Engineering
• Computational Neuroscience
• Machine Learning in Healthcare

Background:

• Parkinson's disease (PD) is a neurodegenerative disorder affecting dopamine production, leading to movement and other symptoms.
• Current PD diagnosis relies on physical/psychological tests, often late in disease progression.
• Voice analysis presents a promising, non-invasive method for early PD detection.

Purpose of the Study:

• To propose and evaluate novel techniques for optimizing machine learning (ML) models for early Parkinson's disease diagnosis using voice analysis.
• To enhance the accuracy and efficiency of PD detection by refining feature selection and hyperparameter tuning of ML algorithms.

Main Methods:

• Utilized voice recordings to extract relevant features for PD diagnosis.
• Applied Synthetic Minority Oversampling Technique (SMOTE) for dataset balancing.
• Employed Recursive Feature Elimination (RFE) for feature selection and t-distributed stochastic neighbour embedding (t-SNE) and principal component analysis (PCA) for dimensionality reduction.
• Trained and evaluated multiple ML classifiers including Support-Vector Machine (SVM), K-Nearest Neighbours (KNN), Decision Tree (DT), Random Forest (RF), and Multilayer Perceptron (MLP).

Main Results:

• The Random Forest (RF) classifier combined with t-distributed stochastic neighbour embedding (t-SNE) achieved 97% accuracy, 96.50% precision, 94% recall, and 95% F1-score.
• The Multilayer Perceptron (MLP) classifier with principal component analysis (PCA) demonstrated superior performance with 98% accuracy, 97.66% precision, 96% recall, and 96.66% F1-score.
• These optimized ML techniques significantly outperformed existing methods for PD diagnosis via voice analysis.

Conclusions:

• Optimized machine learning models, particularly MLP with PCA and RF with t-SNE, offer highly accurate and efficient methods for early Parkinson's disease detection.
• Voice analysis, enhanced by advanced ML techniques, provides a viable and non-invasive approach for diagnosing Parkinson's disease.
• The proposed feature optimization and hyperparameter tuning strategies represent a significant advancement in the early diagnosis of Parkinson's disease.