The recognition accuracy of electrocardiographic electrodes misplacement by an artificial intelligence system compared to medical experts

Background: еlectrocardiographic electrodes misplacement represents one of the most common recording technique errors. Aim: to compare the accuracy of an artificial intelligence software in recognizing misplacement of electrocardiographic electrodes with human analysis by medical experts. Materials and methods: the work was carried out in 3 stages: 1) two electrocardiography datasets with correctly and erroneously positioned electrodes were generated: the training sample (262 electrocardiograms of 27 patients) and the test sample (59 electrocardiograms of 6 patients); 2) artificial intelligence system was trained to correctly annotate electrodes misplacement; 3) testing an artificial intelligence and comparison of its results with the indicators of analysis by doctors-specialists in the field of cardiology. Results: compared to the doctors, the artificial intelligence algorithm functioned more precise and stable in recognizing incorrect electrodes positioning (area under ROC-curve: >0.81 vs 0.58). It also discriminated more error classes (5 of 9 vs. 3 of 9). Doctors and the artificial intelligence system The error classes that were shown to be better determined by the doctors differed from those that were better determined by the algorithm. Conclusion: according to the obtained results, the capability of recognizing incorrect electrode placement demonstrated by the artificial intelligence system is comparable to the diagnostic accuracy of medical analysis. In general, the algorithm’s operation is characterized by greater specificity and stability. It is advisable to improve the algorithm and conduct further research.