Expect to improve patient survival rate by developing an ECG analysis AI model without stopping chest compression
Aug 15, 2024
An artificial intelligence model that can analyze electrocardiogram while continuing chest compression during CPR has been developed by domestic medical staff.
A research team led by Professor Lee Soo-kyo of the Department of Emergency Medicine at Korea University Ansan Hospital (Professor Lee Soo-kyo of the Department of Emergency Medicine and Professor Jeong Soo-min of the Key Research Project) published a research paper on the "Resuscitation Rhythm Analysis of Defibrillation during CPR Using Artificial Intelligence" in the top journal of the Department of Emergency Medicine 'Resuscitation.
In order to increase the survival rate of cardiac arrest patients, it is very important to perform defibrillation as quickly as possible while continuing high-quality CPR. However, before defibrillation, electrocardiogram analysis is essential to identify the heart rhythm that requires defibrillation, that is, the rhythm that requires impact, and in the meantime, chest compression is bound to stop. Chest compression is stopped for up to 10 seconds for electrocardiogram analysis in the hospital and for up to several tens of seconds outside the hospital. The longer the chest compression is interrupted, the lower the survival rate of cardiac arrest patients.
The research team collected 1889 cardiopulmonary resuscitation data from September 2019 to February 2024 at Korea University Ansan Hospital and extracted the rhythm of stopping chest compression. After that, based on the extracted data, the rhythm required for impact and the rhythm was classified and learned to artificial intelligence through a one-dimensional convolutional neural network. One-dimensional convolutional neural networks are deep learning techniques that process data in the form of images, and when applied to medical biosignals, they are more useful, such as enabling real-time data processing.
As a result of learning, it was found that the predictability of the shock-needed rhythm of artificial intelligence was very good. The Area Under Receiver Operating Curve (AUROC) value used as a predictive performance evaluation index is 0.8672, and the closer the AUROC value is to 1, the better the predictive performance is.
Professor Lee said, "This study presents a new paradigm of introducing artificial intelligence technology to CPR and is the world's first study using hospital field data. As we can predict the rhythm that requires defibrillation without stopping chest compression due to electrocardiogram analysis, it will be of great help in reviving patients when commercialized."
Professor Lee's research team plans to collaborate with Seoul National University Hospital and Chonnam National University Hospital to secure more data and conduct follow-up studies to improve prediction accuracy, such as redesigning artificial intelligence. Meanwhile, this study was released from home in June with a domestic patent.
A research team led by Professor Lee Soo-kyo of the Department of Emergency Medicine at Korea University Ansan Hospital (Professor Lee Soo-kyo of the Department of Emergency Medicine and Professor Jeong Soo-min of the Key Research Project) published a research paper on the "Resuscitation Rhythm Analysis of Defibrillation during CPR Using Artificial Intelligence" in the top journal of the Department of Emergency Medicine 'Resuscitation.
In order to increase the survival rate of cardiac arrest patients, it is very important to perform defibrillation as quickly as possible while continuing high-quality CPR. However, before defibrillation, electrocardiogram analysis is essential to identify the heart rhythm that requires defibrillation, that is, the rhythm that requires impact, and in the meantime, chest compression is bound to stop. Chest compression is stopped for up to 10 seconds for electrocardiogram analysis in the hospital and for up to several tens of seconds outside the hospital. The longer the chest compression is interrupted, the lower the survival rate of cardiac arrest patients.
The research team collected 1889 cardiopulmonary resuscitation data from September 2019 to February 2024 at Korea University Ansan Hospital and extracted the rhythm of stopping chest compression. After that, based on the extracted data, the rhythm required for impact and the rhythm was classified and learned to artificial intelligence through a one-dimensional convolutional neural network. One-dimensional convolutional neural networks are deep learning techniques that process data in the form of images, and when applied to medical biosignals, they are more useful, such as enabling real-time data processing.
As a result of learning, it was found that the predictability of the shock-needed rhythm of artificial intelligence was very good. The Area Under Receiver Operating Curve (AUROC) value used as a predictive performance evaluation index is 0.8672, and the closer the AUROC value is to 1, the better the predictive performance is.
Professor Lee said, "This study presents a new paradigm of introducing artificial intelligence technology to CPR and is the world's first study using hospital field data. As we can predict the rhythm that requires defibrillation without stopping chest compression due to electrocardiogram analysis, it will be of great help in reviving patients when commercialized."
Professor Lee's research team plans to collaborate with Seoul National University Hospital and Chonnam National University Hospital to secure more data and conduct follow-up studies to improve prediction accuracy, such as redesigning artificial intelligence. Meanwhile, this study was released from home in June with a domestic patent.
|
bellho@sportschosun.com
