2021 Volume 41 Issue 1 Pages 5-13
Background : Automatic diagnosis of electrocardiograms（ECG）has been widely used not only in clinical practice but in the health check-ups for more than half a century. However, the diagnostic accuracy of widely-used computer equipped electrocardiograph is not always satisfactory and therefore requires an expert’s over-read. The aim of the present study was to evaluate the current situation and factors that led to inappropriate diagnosis focused on atrial fibrillation（AF）. Method : ECGs with inappropriate computer diagnosis were evaluated carefully by members of The Task Force on Automated Diagnosis of the Electrocardiogram in Japan. Results :（1）Inappropriate diagnosis of AF was found in 54 of 50,000 health check-up ECGs（0.11％）and 51 of 272 consecutive patients（18.8％）at a cardiovascular clinic.（2）Factors of 62 under-diagnoses for AF consisted of 42（67.7％）misdiagnoses of f-wave as sinus P waves, 5（8.1％）as ectopic P waves, 8（12.9％）as atrial flutters and 7（11.3％）of no atrial excitations.（3）Factors of 83 over-diagnoses for AF consisted of 38（45.8％）that missed sinus P waves, 37（44.6％）that overlooked ectopic P waves, and 8（9.6％）that misdiagnosed typical flutter waves. Conclusion : Inappropriate automatic diagnosis of AF may result not only in unnecessary consultations with expert physicians due to over-diagnosis but also increased risk of cerebral infarction or heart failure due to under-diagnosis. The construction of highly accurate and reliable systems by further improving computer algorithms is urgent needly to improve automatic diagnosis of ECG.
