Heart failure prognosis risk assessment model based on multimodal data fusion and IoT device monitoring

Heart failure (HF) is a major cardiovascular disease with high global mortality and disease burden. Accurate early prediction and risk assessment are essential but challenging due to HF’s complex pathology, often inadequately assessed using single data sources like imaging or clinical data alone. To enhance HF prognostic accuracy, our study introduces a multimodal data fusion model that integrates patient imaging, clinical features (e.g., age, blood pressure, ejection fraction), and real-time data from IoT devices monitoring physiological parameters. This deep learning framework combines graph neural networks (GNN) and convolutional neural networks (CNN) to extract comprehensive features from diverse data types, thereby improving risk predictions. Tested on Chest X-ray and a proprietary HF electronic medical record dataset, our model demonstrated superior performance in accuracy, AUC, and F1 score compared to traditional methods, especially in predicting death events and identifying high-risk HF patients. This confirms the benefits of incorporating IoT device monitoring into multimodal data fusion for HF prognosis.