Predicting Parkinson's Disease Progression: Evaluation of Ensemble Methods in Machine Learning.

Area of Science:

• Computational Neuroscience
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Parkinson's disease (PD) is a complex neurodegenerative disorder.
• Early and accurate diagnosis of PD is critical for timely intervention and treatment.
• Machine learning (ML) offers promising avenues for improving PD diagnostic accuracy.

Purpose of the Study:

• To conduct a comparative analysis of various machine learning techniques for Parkinson's disease diagnosis.
• To evaluate the effectiveness of clustering and prediction learning approaches in PD diagnosis.
• To identify the most accurate ML method for predicting Motor-UPDRS and Total-UPDRS scores.

Main Methods:

• Utilized clustering techniques (e.g., Expectation-Maximization) for PD data segmentation.
• Employed Support Vector Regression (SVR) ensembles for predicting Motor-UPDRS and Total-UPDRS.
• Compared performance against other methods including Multiple Linear Regression, Neurofuzzy, and standard SVR.

Main Results:

• Expectation-Maximization combined with SVR ensembles demonstrated superior prediction accuracy for Motor-UPDRS and Total-UPDRS.
• This approach outperformed Decision Trees, Deep Belief Networks, Neurofuzzy, and other clustering-SVR combinations.
• The study was validated on a real-world Parkinson's disease dataset.

Conclusions:

• Machine learning, particularly the Expectation-Maximization algorithm with SVR ensembles, significantly enhances PD diagnostic prediction.
• This method offers a more accurate approach for assessing disease progression markers like Motor-UPDRS and Total-UPDRS.
• The findings support the integration of advanced ML techniques into clinical practice for improved PD management.