Prof. Hao-Min Cheng from Taipei Veterans General Hospital and Chair Prof. Vincent S. Tseng from National Yang Ming Chiao Tung University announced the successful development of a heart failure patient readmission and mortality risk prediction technology based on multimodal learning. This technology effectively improves prediction accuracy by integrating multimodal data fusion and deep learning techniques, providing medical professionals with more effective treatment decision-making support.

Automatically extracting effective features for prediction by integrating medical records and multiple imaging modalities

Prof. Hao-Min Cheng stated that heart failure is a serious chronic disease with high rates of readmission and mortality. Traditional prediction methods mainly rely on single tools or basic clinical variables and single biomarkers, making it difficult to accurately reflect the complexity of the disease. The multimodal learning technology developed by the team integrates medical history data, electrocardiograms, chest X-rays, and other data types, and utilizes deep learning techniques to extract effective predictive features, significantly improving prediction accuracy.

In the prediction of readmission and mortality risk in heart failure patients, data such as medical history, electrocardiograms, and chest X-rays reflect different aspects of the patient's condition. Through multimodal data fusion, a more comprehensive understanding of the patient's health status can be achieved. Deep learning techniques automatically extract effective predictive features from multimodal data, reducing the cost and difficulty of manual feature engineering, and further improving prediction accuracy.

Precise prediction! AUC value as high as 0.92

This technology utilizes deep learning techniques to automatically extract effective predictive features, reducing the cost and difficulty of manual feature engineering, and significantly improving prediction accuracy. In predicting readmission and mortality risk in heart failure patients, the AUC values of this technology can reach up to 0.92 and 0.90, demonstrating outstanding predictive capabilities and providing medical professionals with more reliable reference basis.

This technology represents a major breakthrough in the field of heart failure patient prognosis assessment, offering new strategies for the treatment of heart failure patients and improving their prognosis.

Resource (Mandarin):

AI預測心衰竭再入院與死亡風險　準確率高達92%