Joint Coupled-Feature Representation and Coupled Boosting for AD Diagnosis

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Machine Learning

Background:

• Computer-aided diagnosis of Alzheimer's Disease (AD) and Mild Cognitive Impairment (MCI) is an area of significant research interest.
• Existing methods often treat neuroimaging features independently, neglecting the complex interconnected nature of brain regions.

Purpose of the Study:

• To develop a novel approach for AD and MCI diagnosis that accounts for the relationships among neuroimaging features.
• To improve diagnostic accuracy by integrating multi-modal data through a coupled boosting algorithm.

Main Methods:

• Devised a coupled feature representation utilizing intra-coupled and inter-coupled interaction relationships.
• Proposed a novel coupled boosting algorithm for multi-modal data fusion, analyzing pairwise coupled-diversity correlation.
• Introduced a new weight updating function considering both incorrectly and inconsistently classified samples.

Main Results:

• Achieved high accuracies of 94.7% for Alzheimer's Disease (AD) vs. Normal Control (NC) classification.
• Attained 80.1% accuracy for Mild Cognitive Impairment (MCI) vs. Normal Control (NC) classification on the ADNI dataset.
• Outperformed competing and state-of-the-art methods in diagnostic accuracy.

Conclusions:

• The proposed coupled feature representation and boosting algorithm significantly enhance the accuracy of computer-aided AD and MCI diagnosis.
• Considering feature interactions and employing multi-modal data fusion are crucial for robust neurodegenerative disease detection.