User login
PHILADELPHIA – An artificial intelligence platform that sends alerts based on electrocardiography results enabled cardiologists and emergency department physicians at a major hospital in Taiwan to move patients with ST-elevation myocardial infarction (STEMI) into the catheterization laboratory 9 minutes sooner than the conventional protocol that did not use AI.
“This is the first randomized clinical trial to demonstrate the reduction of electrocardiography to coronary cath lab activation time" from 52.3 to 43.3 minutes (P = .003), Chin Sheng Lin, MD, PhD, director of cardiology at the National Defense Medical Center Tri-Service General Hospital in Taipei City, said in presenting the results at the American Heart Association scientific sessions.
Dr. Lin reported results from the Artificial Intelligence Enabled Rapid Identify of ST-Elevation Myocardial Infarction Using Electrocardiogram (ARISE) trial. The trial included 43,994 patients who came to the hospital’s emergency and inpatient departments with at least one ECG but no history of coronary angiography (CAG) in the previous 3 days between May 2022 and April 2023.
They were randomly assigned by date to either AI-ECG for rapid identification and triage of STEMI or standard care. Overall, 145 patients were finally diagnosed with STEMI based on CAG, 77 in the intervention group and 68 in the control group. All patients were seen by one of 20 cardiologists who participated in the study.
Dr. Lin and his group developed an AI algorithm that captures the ECG readout in the emergency department, analyzes the data and then sends a high-risk alarm to the front-line physician and on-duty cardiologist to activate the primary percutaneous coronary intervention (PCI).
Trial results
The differentiation between groups was even more pronounced in ED patients during regular working hours, Dr. Lin said, at 61.6 minutes for the intervention group vs. 33.1 minutes for controls (P = .001).*
He noted that the AI group showed a trend towards fewer cases of clinically suspected STEMI but not getting CAG, 6.5% vs. 15.8%, for an odds ratio of 0.37 (95% confidence interval, 0.14-0.94).
The AI-ECG model also demonstrated a high diagnostic accuracy. “With this AI-ECG system, because it has a very high accuracy and a high positive predictive variable that reach 88%, we can send a message to the on-duty cardiologists and also the emergency room physician and they can send the patients to receive the operation or the PCI as soon as possible,” Dr. Lin said in an interview.
The time differential is critical, Dr. Lin said. “For the patient with acute myocardial infarction, 1 minute is critical, because the patients can die within minutes,” he said. “If we can save 9 minutes I think we can save more lives, but it needs a larger study to evaluate that.”
Dr. Lin acknowledged a few limitations with the trial, among them its single-center nature, relatively small sample size of STEMI patients and the short-term of follow-up. Future study should involve multiple centers along with a prehospital, emergency medical services AI-ECG model.
‘Novel’ for an AI trial
“This is an incredible application of an AI technology in a real-world problem,” said Brahmajee K. Nallamothu, MD, MPH, an interventional cardiologist at the University of Michigan, Ann Arbor, who did not participate in the study. “What I really love about this study is it’s actually a clinical problem that has large implications, particularly for under-resourced areas.”
Using a randomized clinical trial to evaluate the AI platform is “very, very novel,” he said, and called the time improvement “enormous.” Referencing Dr. Lin’s next steps for studying the AI-ECG platform, Dr. Nallamothu said, “if we could push this up even earlier to paramedics and EMTs and prehospital systems, there would be a lot of excitement there.”
He noted the sensitivity analysis resulted in a rate of 88.8% along with the positive predictive value of 88%. “Missing 1 out of 10 ST-elevation MIs in my eyes can still be considered a big deal, so we need to know if this is happening in particular types of patients, for example women versus men, or other groups.”
However, some investigations reported false activation rates as high as 33%, he said. “So, to say that, the positive predictive value is at 88% is really exciting and I think it can make a real inroads,” Dr. Nallamothu said.
Dr. Lin and Dr. Nallamothu have no relevant disclosures.
*Correction, 11/20/23: An earlier version of this article misstated in both trial arms the time to coronary catheterization lab activation.
PHILADELPHIA – An artificial intelligence platform that sends alerts based on electrocardiography results enabled cardiologists and emergency department physicians at a major hospital in Taiwan to move patients with ST-elevation myocardial infarction (STEMI) into the catheterization laboratory 9 minutes sooner than the conventional protocol that did not use AI.
“This is the first randomized clinical trial to demonstrate the reduction of electrocardiography to coronary cath lab activation time" from 52.3 to 43.3 minutes (P = .003), Chin Sheng Lin, MD, PhD, director of cardiology at the National Defense Medical Center Tri-Service General Hospital in Taipei City, said in presenting the results at the American Heart Association scientific sessions.
Dr. Lin reported results from the Artificial Intelligence Enabled Rapid Identify of ST-Elevation Myocardial Infarction Using Electrocardiogram (ARISE) trial. The trial included 43,994 patients who came to the hospital’s emergency and inpatient departments with at least one ECG but no history of coronary angiography (CAG) in the previous 3 days between May 2022 and April 2023.
They were randomly assigned by date to either AI-ECG for rapid identification and triage of STEMI or standard care. Overall, 145 patients were finally diagnosed with STEMI based on CAG, 77 in the intervention group and 68 in the control group. All patients were seen by one of 20 cardiologists who participated in the study.
Dr. Lin and his group developed an AI algorithm that captures the ECG readout in the emergency department, analyzes the data and then sends a high-risk alarm to the front-line physician and on-duty cardiologist to activate the primary percutaneous coronary intervention (PCI).
Trial results
The differentiation between groups was even more pronounced in ED patients during regular working hours, Dr. Lin said, at 61.6 minutes for the intervention group vs. 33.1 minutes for controls (P = .001).*
He noted that the AI group showed a trend towards fewer cases of clinically suspected STEMI but not getting CAG, 6.5% vs. 15.8%, for an odds ratio of 0.37 (95% confidence interval, 0.14-0.94).
The AI-ECG model also demonstrated a high diagnostic accuracy. “With this AI-ECG system, because it has a very high accuracy and a high positive predictive variable that reach 88%, we can send a message to the on-duty cardiologists and also the emergency room physician and they can send the patients to receive the operation or the PCI as soon as possible,” Dr. Lin said in an interview.
The time differential is critical, Dr. Lin said. “For the patient with acute myocardial infarction, 1 minute is critical, because the patients can die within minutes,” he said. “If we can save 9 minutes I think we can save more lives, but it needs a larger study to evaluate that.”
Dr. Lin acknowledged a few limitations with the trial, among them its single-center nature, relatively small sample size of STEMI patients and the short-term of follow-up. Future study should involve multiple centers along with a prehospital, emergency medical services AI-ECG model.
‘Novel’ for an AI trial
“This is an incredible application of an AI technology in a real-world problem,” said Brahmajee K. Nallamothu, MD, MPH, an interventional cardiologist at the University of Michigan, Ann Arbor, who did not participate in the study. “What I really love about this study is it’s actually a clinical problem that has large implications, particularly for under-resourced areas.”
Using a randomized clinical trial to evaluate the AI platform is “very, very novel,” he said, and called the time improvement “enormous.” Referencing Dr. Lin’s next steps for studying the AI-ECG platform, Dr. Nallamothu said, “if we could push this up even earlier to paramedics and EMTs and prehospital systems, there would be a lot of excitement there.”
He noted the sensitivity analysis resulted in a rate of 88.8% along with the positive predictive value of 88%. “Missing 1 out of 10 ST-elevation MIs in my eyes can still be considered a big deal, so we need to know if this is happening in particular types of patients, for example women versus men, or other groups.”
However, some investigations reported false activation rates as high as 33%, he said. “So, to say that, the positive predictive value is at 88% is really exciting and I think it can make a real inroads,” Dr. Nallamothu said.
Dr. Lin and Dr. Nallamothu have no relevant disclosures.
*Correction, 11/20/23: An earlier version of this article misstated in both trial arms the time to coronary catheterization lab activation.
PHILADELPHIA – An artificial intelligence platform that sends alerts based on electrocardiography results enabled cardiologists and emergency department physicians at a major hospital in Taiwan to move patients with ST-elevation myocardial infarction (STEMI) into the catheterization laboratory 9 minutes sooner than the conventional protocol that did not use AI.
“This is the first randomized clinical trial to demonstrate the reduction of electrocardiography to coronary cath lab activation time" from 52.3 to 43.3 minutes (P = .003), Chin Sheng Lin, MD, PhD, director of cardiology at the National Defense Medical Center Tri-Service General Hospital in Taipei City, said in presenting the results at the American Heart Association scientific sessions.
Dr. Lin reported results from the Artificial Intelligence Enabled Rapid Identify of ST-Elevation Myocardial Infarction Using Electrocardiogram (ARISE) trial. The trial included 43,994 patients who came to the hospital’s emergency and inpatient departments with at least one ECG but no history of coronary angiography (CAG) in the previous 3 days between May 2022 and April 2023.
They were randomly assigned by date to either AI-ECG for rapid identification and triage of STEMI or standard care. Overall, 145 patients were finally diagnosed with STEMI based on CAG, 77 in the intervention group and 68 in the control group. All patients were seen by one of 20 cardiologists who participated in the study.
Dr. Lin and his group developed an AI algorithm that captures the ECG readout in the emergency department, analyzes the data and then sends a high-risk alarm to the front-line physician and on-duty cardiologist to activate the primary percutaneous coronary intervention (PCI).
Trial results
The differentiation between groups was even more pronounced in ED patients during regular working hours, Dr. Lin said, at 61.6 minutes for the intervention group vs. 33.1 minutes for controls (P = .001).*
He noted that the AI group showed a trend towards fewer cases of clinically suspected STEMI but not getting CAG, 6.5% vs. 15.8%, for an odds ratio of 0.37 (95% confidence interval, 0.14-0.94).
The AI-ECG model also demonstrated a high diagnostic accuracy. “With this AI-ECG system, because it has a very high accuracy and a high positive predictive variable that reach 88%, we can send a message to the on-duty cardiologists and also the emergency room physician and they can send the patients to receive the operation or the PCI as soon as possible,” Dr. Lin said in an interview.
The time differential is critical, Dr. Lin said. “For the patient with acute myocardial infarction, 1 minute is critical, because the patients can die within minutes,” he said. “If we can save 9 minutes I think we can save more lives, but it needs a larger study to evaluate that.”
Dr. Lin acknowledged a few limitations with the trial, among them its single-center nature, relatively small sample size of STEMI patients and the short-term of follow-up. Future study should involve multiple centers along with a prehospital, emergency medical services AI-ECG model.
‘Novel’ for an AI trial
“This is an incredible application of an AI technology in a real-world problem,” said Brahmajee K. Nallamothu, MD, MPH, an interventional cardiologist at the University of Michigan, Ann Arbor, who did not participate in the study. “What I really love about this study is it’s actually a clinical problem that has large implications, particularly for under-resourced areas.”
Using a randomized clinical trial to evaluate the AI platform is “very, very novel,” he said, and called the time improvement “enormous.” Referencing Dr. Lin’s next steps for studying the AI-ECG platform, Dr. Nallamothu said, “if we could push this up even earlier to paramedics and EMTs and prehospital systems, there would be a lot of excitement there.”
He noted the sensitivity analysis resulted in a rate of 88.8% along with the positive predictive value of 88%. “Missing 1 out of 10 ST-elevation MIs in my eyes can still be considered a big deal, so we need to know if this is happening in particular types of patients, for example women versus men, or other groups.”
However, some investigations reported false activation rates as high as 33%, he said. “So, to say that, the positive predictive value is at 88% is really exciting and I think it can make a real inroads,” Dr. Nallamothu said.
Dr. Lin and Dr. Nallamothu have no relevant disclosures.
*Correction, 11/20/23: An earlier version of this article misstated in both trial arms the time to coronary catheterization lab activation.
AT AHA 2023