Multi-task exclusive relationship learning for alzheimer's disease progression prediction with longitudinal data

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Alzheimer's disease (AD) is a progressive neurodegenerative disorder impacting memory and cognition.
• Early prediction of AD progression is crucial for timely intervention.
• Existing multi-task learning models often overlook the complex associations between different time points in disease progression.

Purpose of the Study:

• To propose a novel multi-task exclusive relationship learning model for estimating clinical measures in Alzheimer's disease.
• To automatically capture intrinsic relationships among different time points using longitudinal imaging data.
• To enhance the accuracy of early-stage Alzheimer's disease progression prediction.

Main Methods:

• Developed a multi-task exclusive relationship learning model incorporating exclusive lasso and relationship-induced regularization.
• Utilized longitudinal imaging data to model intrinsic relatedness across different time points.
• Designed an efficient optimization algorithm to solve the proposed objective function.

Main Results:

• The proposed method effectively selects discriminative features for various time points.
• It successfully models the intrinsic relatedness among different stages of Alzheimer's disease progression.
• Achieved promising performance in cognitive status prediction compared to state-of-the-art methods.

Conclusions:

• The novel model accurately captures temporal relationships in Alzheimer's disease progression.
• It offers improved cognitive status prediction and aids in discovering disease-related biomarkers.
• This approach advances early detection and understanding of Alzheimer's disease.