AI Accurately Detects LV Dysfunction Using Single-Lead Apple Watch ECG

Although work remains to be done before the method is widely adopted, experts see potential for its use as a screening tool.

Artificial intelligence (AI) algorithms trained to interpret single-leading ECGs from the Apple Watch accurately detect signals of left subclinical ventricular systolic function in conceptual studies, opening up the possibility of using measurements Received at home to seek and treat. Patients before they develop a more serious disease.

With an area below the receiver performance curve (AUC) curve of 0.89, the AI-ECG algorithm has a favorable process with other screening tests commonly used in medicine, such as mammography for breast cancer and oral cytology. Cervical Cancer for Cervical Cancer Researchers led by Zachi Attia, PhD (Mayo Clinic, Rochester, MN) noted in a study published online this week in Natural medicine. Some of the results were revealed earlier this year at Heart Rhythm 2022.

Senior author Paul Friedman, MD (Mayo Clinic) told TCTMD that this is “proof of concept” for using AI algorithms to interpret single-leading ECGs from a smartwatch, adding that the work is more valid and more practical. Is necessary before starting the method. More broadly.

He said tests performed at other hospitals were expected to begin in a few days. Next month. “We have planned demonstrations that look incredibly exciting and encouraging, but we must continue – like any new tool – to verify and show that in the real world it improves human life, reduces risk. “It will reduce the risk of death and reduce the risk of disease. And it will require more studies.”

Other experts echoed both calls for more research and a positive outlook for the approach.

Rhodri Davies, MBBS, PhD (University College London, England) said the study was “very fortunate”, citing the strengthening of future design and the use of separate data from those used to download the first algorithm. .

He stated that if the algorithm is validated in future studies and implementation issues are addressed. At that point, the method can find a place to practice treatment. “It should not replace echocardiography or other means of accurately measuring the LV ejection fraction, “But as a screening tool, it has the potential to be really useful.”

Next steps

Researchers have developed AI algorithms that mostly apply to 12 leading ECGs obtained in clinical settings to look for a variety of diseases, including valvular heart disease, hypertrophic cardiomyopathy, amyloid heart disease, electrolyte dysfunction, and Quiet heartbeat. The Mayo Clinic team has previously shown that algorithms trained to look for signs on 12 leading ECGs increase the detection of LV systolic dysfunction by more than 30% compared with conventional care when deployed throughout Primary care practice in Eagle Study.


But so far this method has not been shown to work when the algorithm is modified to interpret the same leading ECGs as what was obtained with the Apple Watch.

It should not replace echocardiography or other means of accurately measuring the fraction of LV discharge, but as a diagnostic tool it is very useful. Rhodri Davies

For the current study, investigators emailed patients who had been treated at the Mayo Clinic and downloaded a mobile patient app to invite them to participate. Recipients are asked to download the Mayo Clinic ECG curriculum, which sends the ECGs measured on the Apple Watch to a safety dashboard that matches the patient’s electronic medical records. Friedman explained that the Apple Watch was chosen because the company made all the raw ECG data available through the Apple Health Kit. The company did not participate in the study.

A total of 3,884 patients were enrolled, of which 2,454 (63%) delivered at least one ECG during the study period. In the latter group, excluded from 46 US states and 11 countries, the average age was 53 years old, 56% female and 88% white. A total of 125,610 ECGs were sent to the safety dashboard, with 92% of patients using the ECG program more than once and half using it more than 5 times. The majority of ECGs (78.5%) were classified as normal sinus rhythm by the Apple Watch, 5.1% as atrial fibrillation and 16.4% undiagnosed.

To evaluate performance in identifying subclinical cardiac dysfunction, the researchers focused on 421 patients who submitted at least one ECG within 30 days of the clinically presented ultrasound examination. Of these, 16 (3.8%) had a rejection fraction of 40% or less, with the majority having few or no symptoms of LV systolic dysfunction.

The AI-ECG algorithm “worked very well” for the detection of cardiac dysfunction, noting that the AUC of 0.89 was similar to that seen with exercise machine tests, Friedman said. AUC 0.85). “An ECG recorded from a consumer device when we perform this AI analysis can detect what could be a life-threatening silent heart condition,” he said.

Also Read :  AI Enters the Verification Stage of Chip Design

Potential use

Friedman called the study the first test of the method, which needed to be followed by further research. But if the findings are valid, it has some potential role for technology in clinical practice. It can first be used to screen for subclinical LV dysfunction in high-risk individuals, such as the elderly or those with diabetes, and then to begin appropriate treatment when it is prescribed. Found.

Another possibility is for patients undergoing chemotherapy for cancer to be able to track heart damage associated with ECGs obtained on a smartwatch, rather than accessing an ultrasound, he said. Maturity.

Partho Sengupta, MD (Robert Wood Johnson University Hospital, Rutgers Robert Wood Johnson Medical School, New Brunswick, NJ) said that the validation of AI-ECG methods for detecting cardiac dysfunction with ECGs The single lead – with a process similar to when the algorithm was applied to 12-lead ECGs – is a “great development” as it brings the approach into the human home.

He raised some questions about the scope of the study, the lack of racial / ethnic diversity among the participants and whether all patients actually had symptoms, and he emphasized the need for research. More made in the center and other countries. With those warnings, he called it a “great technological innovation.” . . . “It democratizes the ability to access that information.”

This is a great promise. Partho Sengupta

Sengupta also sees a role for this approach as a tool to check if the results of the study are released in future research. He suggested that patients with subclinical LV dysfunction could be treated early to prevent the development of heart failure or more closely screened to find the cause of the fraction of discharge that can later be treated. .

He stressed that the potential impact on treatment outcomes was still confirmed in randomized trials. Future studies should also assess whether all patients with congestive heart failure – regardless of the ejection fraction – can be identified using the AI-ECG algorithm, he said.

Implementation Challenges

There are challenges when it comes to implementing such an approach more widely in clinical practice, experts agree. Davies pointed out the need for the infrastructure needed to collect ECG data and for the right physician to examine the information and act on it. He said it makes no sense to inform patients about the fact that their watch is getting signs of LV dysfunction if there is no contact with a cardiologist or other physician who can record the history and work. Imaging tests are necessary to confirm the findings. .

Also Read :  Krach Institute for Tech Diplomacy at Purdue Hosts Inaugural Tech Freedom Awards

Also, as advised, “You have to be careful who you practice it with.” Noting the algorithm will be expected to work better in older patients at higher risk than healthy and younger patients, Davies said. He noted that false positives also need to be reduced.

Friedman agreed that there was a risk of false positives if the screening was performed on the wrong type of person, stating that some operational process was needed to ensure it was delivered to the right people.

Friedman said the technical aspects of the screening have already been built because the safety dashboard has already been built, Friedman said that although most hospitals do not yet have a similar platform, it is likely to get it. In the next few years.

Friedman stressed that while well integrated into the health care system’s electronic medical records system, this strategy should not add too much time for doctors, especially when the hospital is experienced with it, Friedman stressed. . “Once it is in that situation, I do not think it will affect the workflow significantly. It has to be seamless for it to be effective. The doctor was drowned. “We can not make them busier.”

Another question surrounding the implementation of Sengupta pointed out that individuals will suffer from “fatigue of the alarm clock” if they are constantly alerted to possible problems from their smartwatch. . And that is an aspect that needs to be addressed by studying the AI-ECG examination and its impact on clinical outcomes in randomized trials.

Despite the many questions, Sengupta said he was excited about the possibility, especially if the AI-ECG is combined with other wearable monitors in the future.

“This is a very strong promise. This is an important technological step to be achieved.


Leave a Reply

Your email address will not be published.