Real-time Analysis of Massive Continuous Data from a Dialysis Machine to Predict Heart Failure Risk with New Edge AI Platform with NVIDIA
, Chief, Division of Nephrology, Department of Medicine, Taipei Veterans General Hospital
, Senior Software Engineer, Taipei Veterans General Hospital
Hospitalization for heart failure is a serious complication and a major risk factor for death in dialysis patients. This study uses real-time analysis of massive continuous data from the kidney dialysis machine to use machine learning to predict the occurrence of heart failure. Our proposed model uses physiological and real-time dialysis machine values, such as arterial and venous pressure, blood flow rate, etc., and includes multiple data such as patient medical records, test results, medication information, etc., through machine learning algorithms such as decision trees, gradient boosting, and other models to find abnormal patterns for "personal tailored monitoring." A binary classification model was established to predict failure. The results showed 90% accuracy of heart failure prediction.