Identifying Parkinson's Patients: A Functional Gradient Boosting Approach.

Area of Science:

• Neurology
• Artificial Intelligence
• Biomedical Data Science

Background:

• Parkinson's disease (PD) diagnosis is challenging due to subtle, early-stage symptoms.
• Objective identification of PD is crucial for timely intervention and management.
• Existing diagnostic methods may lack sensitivity in early disease detection.

Purpose of the Study:

• To develop and evaluate a machine learning (ML) model for classifying Parkinson's disease (PD) versus healthy controls (HC).
• To leverage features from the Parkinson's Progression Markers Initiative (PPMI) study for PD classification.
• To incorporate domain expert involvement in the feature selection process for enhanced diagnostic accuracy.

Main Methods:

• Utilized a dataset of 1194 participants from the Parkinson's Progression Markers Initiative (PPMI) study.
• Applied machine learning algorithms with a curated set of features.
• Employed domain expert input for feature selection to optimize the classification model.
• Classified subjects into two groups: Parkinson's disease (PD) and Healthy Control (HC).

Main Results:

• Achieved state-of-the-art performance in classifying PD patients from healthy controls.
• Demonstrated high accuracy with minimal feature engineering, suggesting an efficient approach.
• The ML model effectively distinguished between PD and HC groups based on selected features.

Conclusions:

• Machine learning, particularly with expert-guided feature selection, offers a powerful tool for early Parkinson's disease detection.
• The proposed method shows significant promise for improving the diagnostic accuracy of PD.
• This approach, validated on PPMI data, could aid in earlier identification and management of Parkinson's disease.