A novel semi-supervised multi-view clustering framework for screening Parkinson's disease.

Area of Science:

• Medical imaging analysis
• Machine learning in neurology
• Neurodegenerative disease research

Background:

• Parkinson's disease (PD) diagnosis often relies on brain MRI, but current methods face limitations.
• Unsupervised learning struggles with accurate feature extraction from MRIs.
• Deep learning requires extensive data, which is scarce for PD research.
• Existing studies predominantly use single-view MRI data, limiting feature representation.

Purpose of the Study:

• To address the limitations of existing PD diagnosis methods using brain MRI.
• To propose a novel semi-supervised learning framework for improved PD identification.
• To enhance the utilization of both labeled and unlabeled medical MRI data.

Main Methods:

• Developed a Semi-supervised Multi-view learning Clustering architecture (SMC) framework.
• Employed a sliding window method to capture diverse features from MRI data.
• Utilized Linear Discriminant Analysis (LDA) for dimensionality reduction across different feature views.
• Applied single-view and multi-view clustering techniques on processed multi-feature views.

Main Results:

• The proposed SMC framework demonstrated superior clustering performance compared to state-of-the-art unsupervised learning models.
• The multi-view approach effectively leveraged different feature representations from MRI data.
• The method shows promise in enhancing the accuracy of PD identification.

Conclusions:

• The novel SMC framework offers a significant advancement in diagnosing Parkinson's disease using brain MRI.
• This semi-supervised, multi-view approach effectively overcomes data scarcity and feature extraction challenges.
• The findings contribute to more effective identification of PD in real-world clinical settings by integrating labeled and unlabeled MRI data.