Apple researchers have found that AI models not specifically trained on medical data can still estimate heart rate from stethoscope recordings with strong accuracy. In tests using over 20 hours of real hospital heart sounds, Apple’s in-house model, based on CLAP, outperformed traditional methods. Interestingly, shallow and mid-layer representations worked best, and combining AI with classic signal processing improved reliability. While no product plans were announced, the findings hint at future health applications for devices like iPhones, Apple Watches and AirPods.
Source: 9to5Mac
