A novel attention-based cross-modal transfer learning framework for predicting cardiovascular disease.

Area of Science:

• Biomedical Informatics
• Artificial Intelligence in Medicine
• Cardiovascular Research

Background:

• Cardiovascular disease (CVD) is a major global health concern, complicated by genetic, environmental, and lifestyle factors.
• Existing diagnostic methods face challenges in integrating heterogeneous data for effective CVD assessment.
• Early detection and precise treatment of CVD are critical for reducing mortality rates.

Purpose of the Study:

• To introduce and evaluate a novel Attention-Based Cross-Modal (ABCM) transfer learning framework for enhanced CVD prediction.
• To demonstrate the capability of ABCM in merging and analyzing diverse data sources, including clinical, imaging, and genetic information.
• To improve the accuracy and timeliness of cardiovascular disease detection through advanced data integration.

Main Methods:

• Development of an Attention-Based Cross-Modal (ABCM) transfer learning framework.
• Integration of diverse data types: clinical records, medical imagery, and genetic information.
• Utilization of an attention-driven mechanism to identify and prioritize salient features across data modalities.

Main Results:

• The ABCM framework significantly outperformed traditional single-source and other multi-source models in CVD prediction.
• Achieved high predictive performance: 93.5% accuracy, 92.0% precision, 94.5% recall, and 97.2% AUC.
• Demonstrated superior ability to discern intricate interrelationships within heterogeneous patient data.

Conclusions:

• The ABCM framework offers a powerful approach for accurate and early cardiovascular disease detection.
• Cross-modal data integration via attention mechanisms enhances understanding of CVD complexity.
• The model holds significant potential for improving clinical decision-making and personalized patient care in cardiology.