Parkinson's Disease Diagnosis via Joint Learning From Multiple Modalities and Relations

Area of Science:

• Neuroimaging
• Neurology
• Machine Learning

Background:

• Parkinson's disease (PD) is a progressive neurodegenerative disorder impacting motor function.
• Early and accurate diagnosis of PD is essential for timely clinical intervention and managing patient suffering.

Purpose of the Study:

• To develop a joint regression and classification framework for PD diagnosis using multimodal imaging data.
• To implement a multitask feature selection model for identifying disease-related imaging features and predicting clinical scores.

Main Methods:

• Utilized magnetic resonance imaging (MRI) and diffusion tensor imaging (DTI) data.
• Devised a unified multitask feature selection model for joint regression and classification.
• Explored relationships among features, samples, and clinical scores (depression, sleep, olfaction, cognition).

Main Results:

• The multitask framework significantly improved both regression and classification performance for PD diagnosis.
• The proposed method demonstrated superior performance compared to other state-of-the-art techniques.
• Accurate prediction of clinical scores and PD diagnosis labels was achieved.

Conclusions:

• The developed multitask framework offers an effective approach for early and accurate PD diagnosis.
• Computerized predictions provide a quantitative reference for clinical decision support.
• This method enhances the identification of informative, disease-related features from multimodal imaging data.