The team led by Dr. Der-Cherng Tarng, Director of the Department of Internal Medicine at Taipei Veterans General Hospital, has developed the "Real-Time Hemodialysis Artificial Intelligence Prediction System." This innovative system integrates data from dialysis machines, patient health insurance records, and various test results to accurately predict the risk of heart failure in dialysis patients in real-time. It also precisely calculates post-dialysis dry weight, significantly reducing the risk of dialysis-related complications.

Challenges of Traditional Dialysis Monitoring: Delayed Warnings and Dry Weight Inaccuracy

Dialysis patients often face complications due to heart failure or improper dry weight settings. Traditional prediction methods rely heavily on healthcare personnel observing changes in patients' vital signs or dialysis machine alerts to identify complications, which may not provide timely warnings. Additionally, post-dialysis dry weight settings often have errors due to the dynamic changes in body fluid levels, increasing the risk during treatment.

Combining Big Data with Personalized Monitoring for Timely Complication Warnings

The system uses an AIOT (Artificial Intelligence of Things) gateway for dialysis to collect real-time physiological data and data generated by dialysis machines. This information is integrated with health insurance claim data, medical records, test results, and medication information to form a comprehensive personalized health database. Machine learning techniques are then used to continuously train and optimize the heart failure risk prediction model. This model analyzes each patient's physiological data in real-time and determines their current risk of heart failure. When the risk exceeds a preset threshold, the system automatically issues an alert to notify healthcare personnel for early intervention.

Key Advantages of the System:

1. High Accuracy Prediction: With continuous learning and optimization, the model achieves an accuracy of up to 95%, enabling precise prediction of heart failure risks and allowing healthcare providers to take preventive measures early.
2. Personalized Dry Weight Adjustment: Based on the patient's physiological data and changes during dialysis, the system calculates and adjusts the patient's dry weight in real-time. The error in dry weight adjustment is reduced to as low as 40 grams with the help of the AI model.
3. Enhanced Medical Decision-Making Efficiency: The system provides healthcare personnel with an intuitive smart dashboard displaying each patient's real-time physiological data, heart failure risk assessment, and dry weight adjustment recommendations, helping them comprehensively monitor the patient's health status and provide more personalized and accurate medical services.
4. Automated Monitoring: The system automates the monitoring of dialysis patients' physiological data, reducing the workload of healthcare personnel and enabling the timely detection of abnormal conditions, thereby improving the quality of care.

Dr. Der-Cherng Tarng mentioned that the system has already obtained domestic patents and the team is currently in the process of applying for a U.S. patent. The team plans to continue expanding the model parameters to include physiological indicators such as pulmonary edema and hemoglobin to provide more precise risk predictions. They aim to promote this innovative technology through technical collaborations, application program development, and equipment sales.

Resource (mandarin): 北榮血液透析預警系統 即時分析生理數據防腎友心衰！