Classification of Huntington's Disease Stage with Features Derived from Structural and Diffusion-Weighted Imaging

Area of Science:

• Neuroimaging
• Machine Learning
• Neurology

Background:

• Huntington's disease (HD) is a progressive neurodegenerative disorder.
• Early detection and staging are crucial for managing HD.
• Neuroimaging techniques offer potential biomarkers for HD progression.

Purpose of the Study:

• To classify Huntington's disease (HD) stages using machine learning algorithms.
• To evaluate the effectiveness of T1-weighted and diffusion-weighted imaging measures.
• To assess the impact of feature selection and combined imaging modalities on classification accuracy.

Main Methods:

• Support vector machines (SVM) were employed for classification.
• Features were derived from grey matter (GM) and fractional anisotropy (FA) values.
• Feature selection involved whole-brain, region-of-interest (ROI), and automated (Relief-F) approaches.
• Classification was performed on premanifest HD (Pre-HD), early-manifest HD (Early-HD), and healthy control (HC) groups.

Main Results:

• High classification accuracy (85-95%) was achieved between Early-HD and Pre-HD or HC groups.
• FA values from the caudate ROI successfully distinguished Pre-HD from controls with 74% accuracy.
• Combining GM and FA measures did not significantly improve classification performance.

Conclusions:

• Machine learning models can effectively classify HD stages using neuroimaging data.
• Fractional anisotropy (FA) is a sensitive biomarker for early premanifest HD detection.
• Specific ROIs, like the caudate nucleus, are critical for distinguishing early disease stages.