TabulaTime: A Novel Multimodal Deep Learning Framework for Advancing Acute Coronary Syndrome Prediction through Environmental and Clinical Data Integration

TL;DR

TabulaTime presents a multimodal deep learning framework that fuses time-series environmental data with clinical risk factors to improve Acute Coronary Syndrome (ACS) subtype prediction. The key innovation is PatchRWKV, a time-series feature extractor with Patch embedding and an RWKV encoder that processes long sequences with linear complexity, paired with attention-based multimodal integration for accurate and interpretable predictions. Empirical results show TabulaTime outperforms traditional models (e.g., RF, LightGBM, CatBoost) by 20.5%–32.2% in accuracy, with environmental data contributing a 10.1% uplift, and PatchRWKV delivering state-of-the-art performance on classification and forecasting tasks. The model identifies clinically relevant predictors (e.g., Systolic BP, Symptom-to-Admission Time, BMI) alongside environmental factors (PM10, NO, temperature), and demonstrates robustness across periods with varying air quality, underscoring its potential for personalized ACS prevention and informing public health strategies.

Abstract

Acute Coronary Syndromes (ACS), including ST-segment elevation myocardial infarctions (STEMI) and non-ST-segment elevation myocardial infarctions (NSTEMI), remain a leading cause of mortality worldwide. Traditional cardiovascular risk scores rely primarily on clinical data, often overlooking environmental influences like air pollution that significantly impact heart health. Moreover, integrating complex time-series environmental data with clinical records is challenging. We introduce TabulaTime, a multimodal deep learning framework that enhances ACS risk prediction by combining clinical risk factors with air pollution data. TabulaTime features three key innovations: First, it integrates time-series air pollution data with clinical tabular data to improve prediction accuracy. Second, its PatchRWKV module automatically extracts complex temporal patterns, overcoming limitations of traditional feature engineering while maintaining linear computational complexity. Third, attention mechanisms enhance interpretability by revealing interactions between clinical and environmental factors. Experimental results show that TabulaTime improves prediction accuracy by over 20% compared to conventional models such as CatBoost, Random Forest, and LightGBM, with air pollution data alone contributing over a 10% improvement. Feature importance analysis identifies critical predictors including previous angina, systolic blood pressure, PM10, and NO2. Overall, TabulaTime bridges clinical and environmental insights, supporting personalized prevention strategies and informing public health policies to mitigate ACS risk.

Figures (9)

• Figure 1: The flowchart of the proposed TabulaTime framework.
• Figure 2: Architecture of the RWKV Encoder.
• Figure 3: Architecture of the time and channel mixing.
• Figure 4: An illustrative example of using attention mechanisms for multimodal feature integration.
• Figure 5: ROC curve of Catboost(a) and TabulaTime(b) with and without air pollution.