Multi-task fused sparse learning for mild cognitive impairment identification.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Biomedical Engineering

Background:

• Brain functional connectivity networks (BFCN) are vital for understanding brain function and diseases.
• Identifying reliable biomarkers from BFCN is crucial for clinical applications.
• Longitudinal data analysis enhances the understanding of disease progression patterns.

Purpose of the Study:

• To develop a novel longitudinal analysis method for Mild Cognitive Impairment (MCI) classification using resting-state functional magnetic resonance imaging (rs-fMRI).
• To construct biologically meaningful brain networks for improved disease analysis.
• To leverage multi-task learning and fused sparse learning for enhanced network construction.

Main Methods:

• A novel multi-task learning approach integrating fused penalty by regularization was proposed.
• A new objective function for fused sparse learning with smoothness constraints was developed.
• The method was applied to longitudinal rs-fMRI data for MCI classification.

Main Results:

• The proposed method demonstrated high classification performance.
• Achieved 95.74% accuracy for baseline data.
• Achieved 93.64% accuracy for year 1 data.

Conclusions:

• The developed longitudinal analysis method shows promising results for MCI classification.
• The novel multi-task and fused sparse learning approach effectively utilizes brain network data.
• The findings support the potential of this method for early detection and monitoring of MCI.