MLLIFA: Multi-Level Learning and Interactive Fusion Algorithm Combined with Large Foundation Models for Alzheimer's Disease Diagnosis and Etiology Extraction
This algorithm combines imaging and genetic data to enhance Alzheimer's disease diagnosis and understand its causes.
Key Points
The aim is to develop a framework that uses multi-level learning and interactive fusion for diagnosing Alzheimer's disease and extracting its etiology.
Utilized large foundation models to generate high-quality representations of brain region and gene features.
Implemented sparse attention mechanisms for selecting key information from the constructed features.
Applied interaction learning to explore relationships between features within biological contexts.
Achieved an Alzheimer's disease classification accuracy of 91.22%.
Outperformed state-of-the-art methods in diagnosis.
Identified disease-related brain regions, risk genes, and brain-gene associations.