Multimodal Transformer–Based Micronutrient Exposure Profiling for Cardiovascular Event Risk Prediction

DOI

10.2991/978-2-38476-511-9_111How to use a DOI?

Keywords

Multimodal Transformer; Survival Prediction Model; Micronutrient Exposure Profile; Cardiovascular Event Prediction; Multicenter Clinical Data; Risk Calibration and Stratification

Abstract

Cardiovascular disease remains a leading cause of death. Traditional risk equations rely on a small number of clinical variables and struggle to simultaneously capture the combined effects of diet, micronutrients, biochemical markers, genetic susceptibility, and wearable behavioral signals, leading to insufficient discrimination and calibration across different populations. To address this, we propose a multimodal Transformer survival prediction framework for micronutrient exposure profiles. This framework integrates dietary, biochemical, genetic, and wearable time-series information to generate personalized 5-year risk using cross-modal representation learning and time-to-event modeling. This framework is accompanied by posterior calibration and interpretation. Evaluation based on clinical data from four centers demonstrated that our approach achieved a C-index of 0.790/0.782 and an AUROC@5y of 0.842/0.835 on internal and external test sets, with a strong baseline reduction of approximately 9–12% in the integrated Brier score. The net benefit of decision making was higher at thresholds between 5–20%, and the reclassification metrics (NRI ≈ 0.09, IDI ≈ 0.031) supported clinical applicability. Interpretive results showed stable contributions of serum magnesium, 25(OH)D, omega-3 index, folate, and B12, with the emergence of synergistic patterns such as magnesium × omega-3 and folate × B12. This study demonstrates that this paradigm balances discrimination and calibration while maintaining external generalization, providing a translatable technical path for personalized nutritional intervention and risk communication.

Copyright

© 2025 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Tianyao Zhang
PY  - 2025
DA  - 2025/12/31
TI  - Multimodal Transformer–Based Micronutrient Exposure Profiling for Cardiovascular Event Risk Prediction
BT  - Proceedings of the 7th International Conference on Literature, Art and Human Development (ICLAHD 2025)
PB  - Atlantis Press
SP  - 952
EP  - 961
SN  - 2352-5398
UR  - https://doi.org/10.2991/978-2-38476-511-9_111
DO  - 10.2991/978-2-38476-511-9_111
ID  - Zhang2025
ER  -