Multimodal machine learning for 5-year mortality prediction after percutaneous coronary intervention
Byeolhee Kim1,2, Jungyo Suh3, Young-Hak Kim4
1Department of Medical Science, Asan Medical Center, Asan Medical Institute of Convergence Science and Technology, University of Ulsan College of Medicine, Olympic-ro 43-gil, Songpa-gu, Seoul, 05505, Republic of Korea.
Insights
Predicting long-term mortality after percutaneous coronary intervention (PCI) is improved by a new AI model. This multimodal approach integrates imaging, text, and clinical data for better patient outcomes in cardiology.
Area of Science:
- Cardiology
- Artificial Intelligence
- Medical Informatics
Background:
- Percutaneous coronary intervention (PCI) is a key treatment for coronary artery disease.
- Accurate long-term mortality prediction post-PCI is challenging due to complex risk factors.
- Current models often miss prognostic information in imaging and procedural narratives.
Purpose of the Study:
- To develop a multimodal machine learning framework for predicting 5-year all-cause mortality after PCI.
- To integrate coronary angiography video, procedural text, and clinical data for enhanced prediction.
- To establish a foundation for precision medicine in interventional cardiology.
Main Methods:
- A cohort of 10,353 patients undergoing PCI was analyzed.
- Visual embeddings (CLIP) and textual embeddings (BioBERT) were extracted from medical data.
- A trimodal LightGBM model combined visual, textual, and structured clinical features.
Main Results:
- The trimodal model achieved an AUC-ROC of 0.814 for 5-year mortality prediction.
- This significantly outperformed single- and dual-modality models.
- SHAP analysis confirmed complementary prognostic value from unstructured data.
Conclusions:
- Integrating diverse data sources (imaging, text, clinical) significantly improves mortality prediction after PCI.
- Multimodal AI offers a robust and explainable approach for personalized cardiology.
- This framework advances precision medicine by leveraging comprehensive patient information.
Abstract:
Percutaneous coronary intervention (PCI) is a cornerstone treatment for coronary artery disease, yet accurate prediction of long-term mortality remains a critical challenge due to the complex interplay of risk factors. Existing prognostic models rely predominantly on structured clinical data, overlooking the rich, nuanced information embedded in diagnostic imaging and procedural narratives. To address this gap, we present a novel multimodal machine learning framework that integrates coronary angiography video, unstructured procedural text, and structured clinical variables to predict 5-year all-cause mortality. Utilizing a large real-world cohort of 10,353 patients, we extracted visual embeddings via CLIP, textual embeddings via BioBERT, and structured features to construct a unified patient representation. Our trimodal LightGBM model achieved an AUC-ROC of 0.814, significantly outperforming single- and dual-modality baselines ([Formula: see text]). SHAP-based analysis revealed that unstructured data captured complementary prognostic signals, while structured variables provided concentrated predictive strength. This study demonstrates the prognostic value of integrating heterogeneous data sources and establishes a robust, explainable foundation for precision medicine in interventional cardiology.
Related Concept Videos
01:21Cancer Survival Analysis
01:26Coronary Artery Disease IV: Preventive Measures
01:17Classification of Illness
An illness is a response to a disease in which the person's level of functioning is changed compared with a previous level. The general classification of illness includes acute and chronic.
Acute illness is severe...