A team led by Dr. Hsin-Yueh Liang, Director of the Cardiology Department at China Medical University Hospital, has successfully developed an intelligent coronary artery disease detection system. Using a deep learning model, the system can analyze exercise electrocardiogram (ECG) data within just one minute to accurately assess the degree of coronary artery blockage, significantly improving the efficiency and accuracy of heart disease screening.
Accelerated Diagnosis with Deep Learning Model
Traditionally, interpreting exercise ECG results is time-consuming and labor-intensive. Experienced physicians usually need about 3 to 5 minutes to confirm results, while less experienced doctors may take over 15 minutes, increasing the risk of errors. The newly developed system was created through a collaboration between cardiologists, who labeled patient data from those who underwent both exercise ECG and cardiac catheterization within six months, and AI engineers who developed and optimized the algorithm.
The system leverages a deep learning model to analyze over 3,500 data points, incorporating various parameters such as patient age, gender, BMI, maximum heart rate, resting heart rate, maximum heart rate pressure product, maximum workload, maximum ST-segment deviation, and the ST/HR index. It employs Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) models to conduct a comprehensive analysis of the exercise ECG, delivering results within one minute. This aids physicians in making faster and more accurate diagnoses, with an Area Under the Curve (AUC) of 0.882 and an accuracy rate of 0.789, demonstrating its exceptional diagnostic capabilities.
Since its implementation in May 2021, the system has assisted in interpreting over 2,300 exercise ECGs, serving more than 300 patients monthly. The system has been fully integrated into the hospital's internal systems, automatically checking every 60 seconds for new exercise ECG data. This allows clinical physicians to receive AI-assisted reports in real-time during consultations, significantly enhancing diagnostic efficiency. There are also plans to expand the system to other medical institutions to benefit even more patients.
Advancing Smart Healthcare: AI Enhances Medical Quality
Dr. Liang emphasized that in the past, it could take up to two weeks from the initial examination of a patient experiencing chest discomfort to a confirmed diagnosis. With the AI-assisted system, technicians can immediately determine whether a patient has severe coronary artery disease during the examination and recommend a prompt follow-up consultation. This not only reduces the burden on physicians but also allows patients to receive treatment earlier, increasing their chances of survival. The team continues to refine the system, such as by improving data collection methods to minimize interference during exercise ECG monitoring (e.g., body movement, sweating) and by enhancing historical data retrieval through XML to PDF conversion.
Looking ahead, the team plans to further improve the AI model and expand the database to enhance the system's diagnostic capability and generalizability. They will also actively collaborate with other medical institutions to promote this innovative technology, ensuring that more patients benefit from the convenience and efficacy of AI-assisted diagnostics.
Resource (mandarin): 中國附醫AI救心！1分鐘判讀運動心電圖掌握冠狀動脈阻塞程度
