Approved Research

Multi-modal model for prognosis of heart failure

Lay summary

In order to improve the accuracy of heart failure diagnosis and prognosis prediction, it is necessary to analyze complex electronic health record (EHR) data including information on the patient's symptoms, medical history, and social factors, and magnetic resonance imaging (MRI) has been more widely used to predict the prognosis in various heart diseases including heart failure. Therefore, the goal of this study is to develop a multimodal artificial intelligence (AI) system based on EHR and cardiac MRI data to discover new prognostic indicators in the heart and to apply them to patients with heart disease. Existing studies about prognosis prediction of heart failure are solely based on EHR or image data such as MRI. Instead, the multi-modal AI technique based on integrated analysis can yield new clinical indicators by reflecting the complex diagnostic situation of patients with heart failure, and can be applied to individual patient clinical diagnosis by predicting the prognosis for each patient. The proposed project is expected to take 3 years to complete and plans to make a trained models for discovering new clinical indicators and deriving relevance to existing clinical indicators. A new paradigm for the treatment of heart failure can be established by predicting the progression of heart failure patients.