Computer Vision-Guided Automated External Defibrillator System Improves Cardiopulmonary Resuscitation Accuracy: A...

Out-of-hospital cardiac arrests (OHCAs) remain a critical challenge in emergency care, as CPR effectiveness heavily relies on precise chest compressions. We propose a novel CPR guidance system using computer vision to track rescuer motion in real time. This study aims to evaluate the system's efficacy in improving CPR performance, particularly its adaptability to both adult and pediatric populations.

We evaluated a novel CPR navigation system integrated into an Automated External Defibrillator (AED), termed HeartSave M650, which utilizes computer vision to guide CPR quality. A prospective, randomized, crossover trial was conducted comparing HeartSave M650 against the traditional HeartSave M600. In total, 81 volunteers (41 CPR-trained; 40 untrained) stratified into adult (n = 61) and pediatric (n = 20) cohorts were assigned via computer-generated block randomization (block size = 4; stratified by training status) to perform simulated CPR using either: (1) Experimental intervention: HeartSave M650 AED; (2) Control intervention: Conventional HeartSave M600 AED. Participants crossed over after a 15-min washout period. Outcome assessors were masked to device assignment. Participants were not blinded due to visible device differences. The primary outcome was the proportion of compressions within guideline-recommended depth, measured by manikin sensors with video validation.

The primary outcome demonstrated that the M650 significantly improved optimal-depth compressions in adults (60.38% vs. 29.54%, p < 0.001) and pediatric (73.44% vs. 40.96%, p < 0.001), while reducing excessive depth compressions in adults (17.60% vs. 53.97%, p < 0.001) and in pediatric (13.96% vs. 26.11%, p = 0.007).

HeartSave M650 improves compression depth accuracy, more than doubling correct performance compared to conventional AEDs. Its ability to reduce over-depth compressions further supports its clinical value in standardizing CPR quality.