Development of an early diagnosis model for acute heart failure using artificial intelligence deep learning
Feb 03, 2025
A research team led by Joo Hyung-joon and Cha Jung-joon of the Department of Circulatory Mechanics at Korea University Anam Hospital (Professor Joo Hyung-joon and Cha Jung-joon, PhD, Moon Ho-se in the medical information class) has developed a deep learning-based model that can diagnose acute heart failure early based on electrocardiogram conducted in the emergency room. It is attracting great attention from the international community for introducing innovative technology that can quickly and accurately diagnose acute heart failure patients in the emergency room.
Acute heart failure is one of the conditions with a high mortality rate in the emergency room, and accurate early diagnosis can determine the patient's life. However, existing diagnostic methods have many restrictions depending on the patient's condition and test environment, resulting in poor accuracy and speed. The research team introduced artificial intelligence-based electrocardiogram analysis technology to overcome these limitations.
This study was conducted based on ECG data of emergency rooms at Korea University Anam Hospital, Guro Hospital, and Ansan Hospital from 2016 to 2020, and developed a deep learning algorithm by analyzing data from a total of 19,285 patients. The research team extracted key morphological features from electrocardiogram data and combined them with clinical data to compare several machine learning models.
Finally, the model based on the CatBoost algorithm showed the best performance, showing high predictions of 81% accuracy in internal verification and 82% in external verification, demonstrating the best performance. In particular, the model combining ECG data and clinical data had significantly higher diagnostic accuracy than the ECG alone model.
Professor Joo Hyung-jun "This study has opened a new chapter in early diagnosis of acute heart failure"Deep learning models incorporating electrocardiogram and clinical data will overcome the limitations of existing diagnostic methods and enable rapid decision-making in the ER" he explained. Professor Cha Jung-joon added that "if commercialization through the advancement of this technology becomes possible, treatment outcomes for patients with acute heart failure who visit the emergency room can be significantly improved.'
This study is significant in that it has made a technical breakthrough that enables early screening of patients with acute heart failure in an emergency room environment, breaking away from the existing limited diagnostic method. Meanwhile, as a result of the study, 'Deep learning model for identifying acute heart failure patterns using electrocardiography in the emergency room' was published in the latest issue of the European Heart Journal: Acute Cardiovascular Care, an international academic journal on cardiovascular disease.
Acute heart failure is one of the conditions with a high mortality rate in the emergency room, and accurate early diagnosis can determine the patient's life. However, existing diagnostic methods have many restrictions depending on the patient's condition and test environment, resulting in poor accuracy and speed. The research team introduced artificial intelligence-based electrocardiogram analysis technology to overcome these limitations.
This study was conducted based on ECG data of emergency rooms at Korea University Anam Hospital, Guro Hospital, and Ansan Hospital from 2016 to 2020, and developed a deep learning algorithm by analyzing data from a total of 19,285 patients. The research team extracted key morphological features from electrocardiogram data and combined them with clinical data to compare several machine learning models.
Finally, the model based on the CatBoost algorithm showed the best performance, showing high predictions of 81% accuracy in internal verification and 82% in external verification, demonstrating the best performance. In particular, the model combining ECG data and clinical data had significantly higher diagnostic accuracy than the ECG alone model.
Professor Joo Hyung-jun "This study has opened a new chapter in early diagnosis of acute heart failure"Deep learning models incorporating electrocardiogram and clinical data will overcome the limitations of existing diagnostic methods and enable rapid decision-making in the ER" he explained. Professor Cha Jung-joon added that "if commercialization through the advancement of this technology becomes possible, treatment outcomes for patients with acute heart failure who visit the emergency room can be significantly improved.'
This study is significant in that it has made a technical breakthrough that enables early screening of patients with acute heart failure in an emergency room environment, breaking away from the existing limited diagnostic method. Meanwhile, as a result of the study, 'Deep learning model for identifying acute heart failure patterns using electrocardiography in the emergency room' was published in the latest issue of the European Heart Journal: Acute Cardiovascular Care, an international academic journal on cardiovascular disease.
|
This article was translated by Naver AI translator.
