TL;DR
Get workout gear delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
A study presented at the American Heart Association meeting found that an AI model applied to Apple and Samsung smartwatches could distinguish reduced left ventricular function in a clinical cohort, but readings often disagreed between devices and within the same person. Researchers say the findings do not yet establish how well smartwatch screening would work in routine care or in people without diagnosed heart failure.
A Korean study presented at the American Heart Association annual meeting found that an AI model using smartwatch ECGs could identify reduced left ventricular systolic function, but results were not consistent across Apple and Samsung devices or across repeated home readings. The findings raise questions about using smartwatch alerts as a dependable screening test for left ventricular systolic dysfunction (LVSD), a condition in which the heart’s main pumping chamber has reduced function.
The researchers studied 2,211 people at Incheon Sejong Hospital who were undergoing echocardiography for clinical reasons. Participants used both an Apple Watch and a Samsung Galaxy Watch for ECG recordings. Against the study’s definition of LVSD—an ejection fraction of 40% or less—the model’s area under the curve was 0.907 for Apple and 0.895 for Samsung. Sensitivity was 69.2% and 53.4%, respectively, while specificity was 88.9% and 93.4%.
Those overall performance measures did not mean the devices agreed on individual patients. At the same prespecified model cutoff, 9% of the cohort received discordant results: 158 people were positive only on Apple Watch and 42 only on Galaxy Watch. Presenter Hak Seung Lee said Apple Watch scores tended to be higher. In a separate group of 146 people who recorded ECGs at home, 49.3% had both positive and negative results in the same week at the selected cutoff, using one smartwatch, usually a Galaxy Watch.
The model, AiTiALVSD-1L, was trained first on nearly 3 million unlabeled ECGs and then fine-tuned using about 360,000 clinical lead I ECGs, Lee said. The researchers set a cutoff using a separate Galaxy Watch tuning group and applied it to both watch brands. Attempts to improve consistency, including averaging readings over a week, had limited success: stricter criteria reduced false positives but also missed more cases of LVSD.
Why Mixed Watch Readings Matter
The results matter because smartwatch ECGs could offer a more accessible way to flag people who may have impaired heart function. But a screening alert can lead to follow-up testing and clinical workload, while a negative reading could give false reassurance if the test misses a case. The study’s device disagreement and week-to-week variability show why performance cannot be judged only by an aggregate accuracy measure.
At a press conference, University of Michigan cardiologist Jessica Golbus said health systems need to balance false-positive follow-up against identifying people with symptomatic or asymptomatic LVSD. She said systems also need a plan to connect people who receive positive results with clinical evaluation and appropriate treatment. The study did not establish whether using these watch readings improves diagnosis, treatment, or patient outcomes.
Products Worth Considering
Withings BPM Connect Blood Pressure Monitor with Wi-Fi and Bluetooth
Garmin Forerunner 935 GPS Running and Multisport Watch, Black
Garmin Forerunner 165 GPS Running Smartwatch, 43 mm, Whitestone
As an affiliate, we earn on qualifying purchases.
From Promising Model to Screening Test
Interest in smartwatch detection has been encouraged by earlier research. A 2022 study reported good accuracy for an AI model using a smartwatch’s single-lead ECG to detect LVSD. The present study tested a related approach in two prospective cohorts at one hospital in South Korea, including people already referred for echocardiography.
In the device-comparison cohort, 6.6% had known LVSD and 31% had known heart failure. The separate home-recording cohort had a higher-risk profile: 33.6% had LVSD and 55.5% had known heart failure. These populations are not the same as people with no known heart condition seeking screening through a consumer watch, so the findings cannot establish performance in that broader group.
“A single negative home recording is not reassuring on its own, and mixed results should prompt further evaluation.”
— Hak Seung Lee, study presenter
Where the Evidence Falls Short
The research was conducted at one hospital, and the investigators did not test screening in asymptomatic people without heart failure. The study therefore does not show how the model would perform in a general smartwatch-owning population, where LVSD may be less common and the balance between false alerts and missed cases could differ.
It is also unclear how performance varies by age, race, ethnicity, or ejection fraction, and whether results would hold across other devices, software versions, and clinical settings. The reported measures describe this study’s cohorts; they do not establish that a smartwatch reading can diagnose or rule out LVSD for an individual.
Testing Devices in Real Care
Further research will need to assess device-specific cutoffs and whether repeated readings can improve reliability without increasing missed cases. Researchers also need to study people with different clinical backgrounds and those without known heart failure, then examine whether smartwatch alerts lead to timely assessment and better clinical outcomes.
Golbus said health systems will need follow-up pathways for people with positive readings and ways to manage false positives. Until those questions are addressed, the study supports treating smartwatch AI-ECG results as signals that may warrant clinical evaluation, not as a stand-alone diagnosis or a reason to dismiss concern after one negative recording.
Key Questions
What condition was the smartwatch model designed to detect?
It was designed to flag left ventricular systolic dysfunction, or reduced pumping function of the heart’s left ventricle. The study defined it as an ejection fraction of 40% or less.
Did the Apple Watch and Galaxy Watch always agree?
No. At the study’s shared cutoff, 9% of the 2,211-person cohort had a positive result on one watch but not the other.
Can one negative smartwatch ECG rule out LVSD?
The study does not support using a single negative home recording to rule out LVSD. Presenter Hak Seung Lee said one negative reading was not reassuring on its own; people with health concerns should seek guidance from a clinician.
Does this study show that smartwatch screening improves health outcomes?
No. The study examined model performance and consistency, not whether smartwatch screening improves diagnosis, treatment, or patient outcomes.
Who was included in the study?
The main comparison included 2,211 people referred for clinically indicated echocardiography at one South Korean hospital. A separate group of 146 people recorded ECGs at home; both groups included substantial numbers of people with known heart failure.
Source: medpagetoday.com
