Maximum Mean Discrepancy Based Multiple Kernel Learning for Incomplete Multimodality Neuroimaging Data.

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Alzheimer's Disease (AD) diagnosis is challenging with incomplete multimodal data.
• Existing methods like low-rank matrix completion and multi-task learning fail to capture complex relationships and ignore data heterogeneity.

Purpose of the Study:

• To propose a novel Maximum Mean Discrepancy (MMD) based Multiple Kernel Learning (MKL) method for AD diagnosis using incomplete multimodal data.
• To address the limitations of current methods in modeling complex data-to-label relationships and handling modality heterogeneity.

Main Methods:

• Developed a new MMD algorithm to map samples from different modalities into a Reproducing Kernel Hilbert Space (RKHS).
• Integrated data distribution matching, pair-wise sample matching, and feature selection into a unified formulation.
• Utilized MRI and PET data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset.

Main Results:

• The proposed MMD-based MKL method effectively handles modality heterogeneity, making samples comparable for a shared classifier.
• The method successfully captures nonlinear data-to-label relationships crucial for accurate AD diagnosis.
• Experimental results demonstrated superior performance compared to existing methods.

Conclusions:

• The novel MMD-based MKL approach offers a robust solution for Alzheimer's Disease diagnosis with incomplete multimodal data.
• This method improves diagnostic accuracy by effectively managing data heterogeneity and complex relationships.
• The findings suggest a promising direction for advancing AI-driven diagnostic tools in neurodegenerative diseases.