Artificial Intelligence Tool Accurately Diagnoses Structural Heart Disease Using Single-Lead ECGs on Smartwatches

Strategic Insights -

Researchers have made a significant breakthrough in the field of cardiology by developing an artificial intelligence (AI) tool that can accurately diagnose structural heart disease using single-lead electrocardiograms (ECGs) on smartwatches. The AI algorithm, tested on a group of 600 adults, demonstrated a high level of accuracy in detecting structural heart diseases such as weakened pumping ability, damaged valves, or thickened heart muscle.

Key Takeaways:

  • The AI algorithm was developed using a database of 266,054 ECGs from 110,006 patients who received testing and treatment at Yale New Haven Hospital between 2015 and 2023.
  • The algorithm was able to accurately identify structural heart disease from single-lead ECGs obtained from smartwatches with a performance of 92% in a simulated setting and 88% in a real-world setting.
  • The AI tool was able to detect structural heart disease with a sensitivity of 86% and a negative predictive value of 99%.
  • The researchers added "noise" to the model during training to make it more resilient and reliable when dealing with less-than-perfect signals.
  • The tool was prospectively tested on 600 patients, with a median age of 62 years, and was found to be accurate in detecting structural heart disease in about 5% of the participants.
  • The researchers plan to evaluate the AI tool in broader settings and explore its potential integration into community-based heart disease screening programs.

Statistics:

  • 600 participants were tested with the AI tool, with a median age of 62 years.
  • 44% of participants were non-Hispanic white, 15% non-Hispanic Black, 7% Hispanic, 1% Asian, and 33% others.
  • The AI tool accurately detected about 5% of the participants with structural heart disease.
  • The accuracy of the AI tool in a simulated setting was 92%, while in a real-world setting, it was 88%.
  • The sensitivity of the AI tool was 86%, and its negative predictive value was 99%.

Sources:

*

*

*

*

*

*

*

*