This project investigates the classification of patients with heart diseases from healthy controls based on multimodal data (Fig. 11A). The set of multimodal data includes: age, heartbeat (BPM), high blood pressure, low blood pressure, glucose concentration, KCM, and troponin. The project infers whether a person had a heart attack or not using these data. It explores multiple machine learning methods, including SVM, Gaussian Process, MLP and Deep Neural Network (DNN). Most models perform well with high accuracy above 90% (Fig. 11B). Post-hoc SHAP analysis reveals that troponin and KCM features contribute substantially to inference (Fig. 11C).