Tensor Completion for Alzheimer's Disease Prediction From Diffusion Tensor Imaging.

Area of Science:

• Neuroscience
• Medical Imaging Analysis
• Machine Learning

Background:

• Alzheimer's disease (AD) is a progressive neurodegenerative disorder.
• Accurate prediction of AD progression is crucial for patient management.
• Existing methods often overlook the structural information within cognitive scores.

Purpose of the Study:

• To propose a novel method for predicting Alzheimer's disease progression.
• To leverage the inherent structure of cognitive scores for improved prediction.
• To address limitations in current prediction models.

Main Methods:

• Developed a relation-aware tensor completion multitask learning (RATC-MTL) method.
• Represented cognitive scores as a third-order tensor to preserve global structure.
• Integrated tensor completion with linear regression to model score correlations and biological features.

Main Results:

• The proposed RATC-MTL method achieved superior performance in predicting cognitive scores compared to single-task and state-of-the-art multi-task algorithms.
• The method demonstrated the best overall performance in classifying Alzheimer's disease progression.
• Results confirmed the framework's effectiveness in utilizing global cognitive score structure.

Conclusions:

• The developed framework effectively explores global structure information in cognitive scores.
• Tensor completion offers a novel approach for disease diagnosis, particularly in data-scarce scenarios.
• This method enhances the prediction of Alzheimer's disease progression.