Hypergraph-regularized multimodal learning by graph diffusion for imaging genetics based Alzheimer's Disease...

Area of Science:

• Neuroscience
• Medical Imaging
• Genetics

Background:

• Multi-modal data fusion is crucial for diagnosing complex brain disorders.
• Existing methods often use simple feature combinations, neglecting imaging-gene associations.

Purpose of the Study:

• To develop a novel method (HMGD) for joint association learning and outcome prediction using multi-modal data.
• To improve the diagnosis and prediction of brain disorders by integrating imaging and genetic information.

Main Methods:

• A graph diffusion method enhances subject similarity across multi-modality phenotypes.
• Hypergraph regularization incorporates inter- and cross-modality information to identify imaging phenotypes linked to risk single nucleotide polymorphisms (SNPs).
• A multi-kernel support vector machine (MK-SVM) fuses selected phenotypic features for diagnosis and prediction.

Main Results:

• The proposed HMGD approach demonstrated superior performance compared to competing algorithms on Alzheimer's Disease Neuroimaging Initiative (ADNI) datasets.
• HMGD identified significant, consistent, and robust regions of interest (ROIs) linking imaging phenotypes to genetic risk biomarkers.
• The method achieved strong associations for disease interpretation and prediction.

Conclusions:

• HMGD offers an effective framework for multi-modal data fusion in brain disorder research.
• The approach successfully integrates diverse data sources to uncover complex relationships between imaging and genetics.
• This method holds promise for advancing the diagnosis and prediction of neurological conditions.