Adaptive sparse learning using multi-template for neurodegenerative disease diagnosis.

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Neurodegenerative diseases impact millions, particularly the elderly, necessitating early detection.
• Clinical symptoms appear years after neurodegeneration onset, highlighting the need for advanced diagnostic tools.

Purpose of the Study:

• To propose an adaptive feature learning framework for the early diagnosis of neurodegenerative diseases.
• To develop a multi-classification scheme using brain parcellation atlases and predict disease severity.

Main Methods:

• Utilized multiple templates for adaptive feature learning and a multi-classification scheme with various regions of interest.
• Extracted, fused, and selected features with an adaptive sparse degree, integrating linear discriminative analysis and locally preserving projections.
• Constructed a least square regression model and a feature space to predict disease severity guided by clinical scores.

Main Results:

• Validated the framework on Alzheimer's Disease Neuroimaging Initiative and Parkinson's Progression Markers Initiative databases.
• Demonstrated superior performance compared to state-of-the-art methods like multi-modal multi-task learning and joint sparse learning.
• Accurate feature learning identified highly relevant brain regions contributing significantly to disease progression prediction.

Conclusions:

• The proposed adaptive feature learning framework enables accurate early diagnosis and severity prediction of neurodegenerative diseases.
• The method effectively identifies critical brain regions, advancing medical analysis and practical diagnostic applications.
• This approach offers a promising tool for improving patient outcomes through timely intervention.