Multimodal Classification Framework Based on Hypergraph Latent Relation for End-Stage Renal Disease Associated with...

Area of Science:

• Neuroimaging
• Medical Diagnostics
• Artificial Intelligence

Background:

• End-stage renal disease with mild cognitive impairment (ESRDaMCI) presents challenges in diagnosis due to subtle brain network alterations.
• Current multimodal classification methods often fail to capture complex brain region interactions and can be affected by feature noise.

Purpose of the Study:

• To develop an advanced multimodal classification framework for improved recognition of ESRDaMCI.
• To address limitations in existing methods by incorporating high-order relationships and reducing noise in feature matrices.

Main Methods:

• A novel framework utilizing hypergraph latent relation (HLR) was proposed.
• Functional magnetic resonance imaging (fMRI) data was used to construct a brain functional network with hypergraph structural information.
• Arterial spin labeling (ASL) derived cerebral blood flow (CBF) served as a second modal feature.
• Latent relation adaptive similarity learning (LRAS) was employed for multimodal feature selection (LRMFS).

Main Results:

• The proposed LRMFS framework achieved a classification accuracy (ACC) of 88.67%.
• This represents an improvement of at least 2.84% compared to state-of-the-art methods.
• The framework effectively preserved valuable inter-regional brain information and reduced noise.

Conclusions:

• The developed multimodal classification framework offers a promising approach for identifying imaging markers of ESRDaMCI.
• This method enhances the understanding of brain network properties in ESRDaMCI.
• The findings provide a valuable reference for clinical ESRDaMCI recognition.