Article Details

Abstract

Cardiovascular disease is a leading cause of death worldwide, and early accurate assessment is crucial. Traditional methods rely on limited indicators and struggle to integrate multi-source heterogeneous data. A health assessment model based on clinical-guided cross-modal attention fusion is proposed. It uses clinical data as a query to dynamically guide multimodal fusion, and combines with temporal modeling and uncertainty quantification, to achieve accurate risk prediction. Using data from 3 200 cases in the UK Biobank, the model achieves an area under the receiver operating characteristic curve of 85.6%, a 3.5 percentage point improvement over the optimal baseline. Ablation studies show that removing this mechanism reduces the area under the curve by 2.8%. Uncertainty quantification reveals that the error rate for high-uncertainty samples is 4.7 times that of low-uncertainty samples. This model provides a new interpretable and trustworthy approach for intelligent cardiovascular disease risk assessment.

Keywords

cardiovascular disease; risk prediction; multimodal fusion; Attention Mechanism; clinical guidance; uncertainty quantification

References

1. [1] World Health Organization. Cardiovascular Diseases [EB/OL].[2026-03-11].https://www.who.int/health-topics/cardiovascular-diseases/#tab=tab_1.
2. [2] Liu Mingbo, Wang Zengwu, Fan Jing, etc Key points interpretation of the "China Cardiovascular Health and Disease Report 2023" [J]. Chinese Journal of Cardiovascular Disease Research, 2024, 22 (7): 577-593
3. [3] DAMEN J A A G，HOOFT L，SCHUIT E，et al. Prediction Models for Cardiovascular Disease Risk in the General Population: Systematic Review [J/OL].BMJ，2016，353：i2416（2016-05-16）[2026-01-17].https://pubmed.ncbi.nlm.nih.gov/27184143/.
4. [4] AL-KINDI S，NASIR K. From Data to Wisdom: Harnessing the Power of Multimodal Approach for Personalized Atherosclerotic Cardiovascular Risk Assessment [J].European Heart Journal-Digital Health，2024，5（1）：6-8.
5. [5] KLINE A，WANG H Y，LI Y K，et al. Multimodal Machine Learning in Precision Health: A Scoping Review [J/OL].NPJ Digital Medicine，2022，5（1）：171（2022-11-07）[2026-01-17].https://www.nature.com/articles/s41746-022-00712-8.
6. [6] ZHOU Z Q，WANG J W. A Narrative Review of Multimodal Data Fusion Strategies for Precision Risk Prediction in Coronary Artery Disease: Advances, Challenges, and Future Informatics Directions [J/OL].Rambam Maimonides Medical Journal，2025，16（4）：e0023[2026-01-17].https://www.rmmj.org.il/issues/68/articles/4008.
7. [7] JOHNSON K W，SOTO J T，GLICKSBERG B S，et al. Artificial Intelligence in Cardiology [J].Journal of the American College of Cardiology，2018，71（23）：2668-2679.
8. [8] DUAN J W，XIONG J Q，LI Y H，et al. Deep Learning Based Multimodal Biomedical Data Fusion: An Overview and Comparative Review [J/OL].Information Fusion，2024，112：102536[2026-01-17].https://doi.org/10.1016/j.inffus.2024.102536.
9. [9] LAI C X，YIN M L，KHOLMOVSKI E G，et al. Multimodal AI to Forecast Arrhythmic Death in Hypertrophic Cardiomyopathy [J].Nature Cardiovascular Research，2025，4（7）：891-903.
10. [10] ABDAR M，POURPANAH F，HUSSAIN S，et al. A Review of Uncertainty Quantification in Deep Learning: Techniques, Applications and Challenges [J].Information Fusion，2021，76：243-297.
11. [11] LUNDBERG S M，LEE S-I. A Unified Approach to Interpreting Model Predictions [C]//31st Conference on Neural Information Processing Systems.Long Beach：NIPS，2017：4765-4774.
12. [12] BYCROFT C，FREEMAN C，PETKOVA D，et al. The UK Biobank resource with Deep Phenotyping and Genomic Data [J].Nature，2018，562（7726）：203-209.

About the authors

Wu Jichuan (1996-), male, Han ethnicity, from Anqing, Anhui Province, is currently pursuing a master's degree in multimodal deep learning.