Identifying Imaging Markers for Predicting Cognitive Assessments Using Wasserstein Distances Based Matrix Regression.

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Alzheimer's disease (AD) is a progressive neurodegenerative disorder impacting memory and cognition.
• Early detection and diagnosis of AD rely on identifying neuroimaging markers and predicting cognitive function.
• Current regression models often overlook the intrinsic geometry of cognitive data.

Purpose of the Study:

• To propose a novel robust matrix regression model for Alzheimer's disease (AD) research.
• To integrate Wasserstein distances into regression models to capture the latent geometry of cognitive data.
• To improve the prediction of cognitive assessments and identification of neuroimaging markers for AD.

Main Methods:

• Development of a robust matrix regression model incorporating Wasserstein distances in both the loss function and regularization.
• Introduction of an efficient algorithm for solving the proposed model, including convergence analysis.
• Application and evaluation of the model using cognitive data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort.

Main Results:

• The proposed model effectively integrates Wasserstein distance, uncovering the underlying geometry of cognitive data.
• Empirical results on the ADNI cohort demonstrate the method's high effectiveness.
• The approach shows significant promise for clinical cognitive prediction in Alzheimer's disease.

Conclusions:

• The novel Wasserstein distance-based regression model offers a more geometrically aware approach to analyzing cognitive data in Alzheimer's disease.
• This method enhances the prediction of cognitive decline and the identification of key neuroimaging biomarkers.
• The findings support the utility of this advanced regression technique for early AD detection and clinical applications.