WATCH-LV
2026
Consumer Smartwatch Artificial Intelligence-Enabled Electrocardiogram for Left Ventricular Systolic Dysfunction: Cross-Manufacturer Validation and Serial Home Monitoring (WATCH-LV)
In adults referred for echocardiography, an AI model reading single-lead smartwatch ECGs distinguished people with and without left ventricular systolic dysfunction about equally well on two brands of watch, although 9% of results disagreed between devices.
Résultat
Critère principal: Diagnostic performance for left ventricular systolic dysfunction
The AI model distinguished people with and without LVSD about equally well on both watches. At the same AI-score cutoff, 200 participants (9%) were classified as positive on one watch and negative on the other. In the home monitoring cohort, requiring at least 2 of 3 consecutive positive recording days reduced false-positive classifications from 18 to 9 among the 70 participants without LVSD, and identified 31 of 38 participants with LVSD compared with 33 using single-positive days. 81% of those with consistently positive results had LVSD confirmed by echocardiography vs about 6% of those with consistently negative results. Nearly half (49%) had mixed results, and 40% of this group had confirmed LVSD.
- Sous-thème
- Heart failure & cardiomyopathy
- Schéma
- Validation study of an AI algorithm reading single-lead consumer-smartwatch ECGs from several manufacturers for left ventricular systolic dysfunction, with serial home recordings
- Inclus
- 2 211 patients
- Source
- Lee HS. Presented at AHA Scientific Sessions 2026, 9 November. American Heart Association newsroom, late-breaking abstract 27869. Not yet published.
Discussion et critique
This single-centre study suggests an AI model can flag left ventricular systolic dysfunction from single-lead smartwatch ECGs on two manufacturers' devices, but 9% of participants got discordant results between watches. The release reports no area under the curve, sensitivity or specificity for the clinic cohort, and the population was referred for echocardiography, so performance in an unselected screening population is unknown. The serial home monitoring analysis included only 108 participants and is exploratory: poor-quality recordings were excluded and participants were reminded to record. The study did not test whether acting on the results leads to earlier treatment or better clinical outcomes, and the abstract has not been peer reviewed.