[Sportschosun Reporter Jang Jong-ho] A study has found that analyzing a patient's preoperative electrocardiogram with artificial intelligence (AI) can identify patients at high risk of death after surgery and effectively screen out those who are unlikely to need advanced cardiac testing.
A research team led by Professor Choi Hong-mi, resident Kim Ye-rin, and Professor Jo Young-jin of the Cardiology Department at Seoul National University Bundang Hospital, together with Professor Song In-ae of the Department of Anesthesiology and Pain Medicine, analyzed data from 46,000 non-cardiac surgeries performed between 2020 and 2021 to evaluate the AI ECG's ability to predict risk and determine who should undergo further testing.
An electrocardiogram records the heart's electrical signals as waveforms, but subtle differences and complex patterns are difficult for the human eye to interpret fully. Professor Kim Joong-hee of Emergency Medicine (EM) and Professor Jo Young-jin of Cardiology at Seoul National University Bundang Hospital developed ECG Buddy, the AI solution now most actively used in emergency rooms nationwide for ECG analysis, after recognizing that AI could quantify risks that conventional readings might miss.
In this study, the team applied several heart-disease-related indicators provided by the solution to preoperative patient assessment. Because every surgery can place stress on the cardiovascular system during anesthesia and the procedure itself, accurately assessing that risk beforehand is essential.
First, the researchers used the AI severity score, or QCG-Critical score, to evaluate how accurately it could predict death within 30 days after surgery. The results showed a clear correlation between the AI-predicted risk score and mortality: the death rate was 0.1% in the group with scores below 10, but rose to 11.7% in the group with scores above 40.
The AUROC for predicting 30-day postoperative mortality was 0.909, which is higher than international standard methods such as the European Society of Cardiology assessment model (0.728) and the cardiac risk index RCRI (0.725). It also slightly outperformed the ASA classification, which evaluates a patient's overall physical condition, at 0.886.
The team then examined whether AI analysis of preoperative ECGs could efficiently identify patients who needed advanced tests such as echocardiography or coronary CT. By combining eight AI ECG biomarkers related to cardiac function and disease, they classified patients into a low-risk group, with all indicators within the normal range, and a high-risk group, with at least one abnormal finding.
The analysis found that 92.3% of all surgeries fell into the AI ECG low-risk group, and in those cases, the rate of death or emergency coronary intervention within 30 days after surgery was just 0.2%. Among these patients, 91% of preoperative advanced test results were actually normal, while the cost of those tests accounted for 62.8% of all preoperative cardiovascular advanced testing. The findings suggest that introducing the solution could significantly reduce the burden of preoperative testing.
Notably, all of the results were obtained solely by analyzing a single ECG taken before surgery, without inputting complex data such as laboratory values. The team explained that "given the low cost and simplicity of ECG testing, its value in clinical settings is very high."
For now, however, the researchers said AI ECG should be seen not as a full replacement for existing preoperative advanced testing, but as a screening gate that helps identify high-risk patients and reduce the overall testing burden.
Professor Jo Young-jin of Seoul National University Bundang Hospital said, "AI-based ECG analysis technology is already being actively used in emergency settings as a screening tool before advanced testing," adding, "Because it is fast, simple, and inexpensive, it can also be used in routine surgical care to reduce testing burdens and identify high-risk patients."
Meanwhile, the study's findings on the accuracy and clinical value of AI ECG analysis technology were published in the prestigious international journals European Heart Journal-Digital Health and Journal of Medical Internet Research, respectively.
Reporter Jang Jong-ho bellho@sportschosun.com
This article has been translated by GripLabs Mingo AI.
