Abstract
ECG diagnosis by a computer based on median value was evaluated. 1447 subjects were randomly selected. A cardiologist diagnosed 1284 cases as normal and 173 as abnormal. Out of the 1284 normal cases this computer analysis correctly diagnosed 1225 (95.4%) as normal and 59 (4.6%) as abnormal. Among the 59 cases, 42 were interpreted as myocardial infarction (MI), 8 as myocardial damage (MD), 8 as left ventricular hypertrophy (LVH), and one as incomplete right bundle branch block (IRBBB). Out of 173 abnormal cases, it also correctly diagnosed 159 (92%) as abnormal. But it erroneously diagnosed 14 cases (8%), among which a cardiologist diagnosed 7 as MD, 3 as the first degree atrio-ventricular block, 2 as IRBBB, one as LVH, and another one as PQ shortening. The overall similarity between the diagnoses by a cardiologist and by the computer using a median value analysis was established. As the second attempt, 81 cases were randomly selected from the 1225 normal cases, from which 421 beats were obtained. Diagnoses from 389 beats (92%) were normal which were consistent with those by the median value analysis, while 32 beats (8%) were erroneously diagnosed as MD (11 beats), MI (18 beats), LVH (3 beats). 15 cases were also selected from the 159 abnormal cases from which 76 beats were obtained. 68 beats (89.5%) were correctly diagnosed as MD (29 beats), MI (13 beats), IRBBB (4 beats), and questionable MI (16 beast) but 8 beats (10.5%) were erroneously diagnosed as normal. Therefore, it can be concluded that the computer diagnosis of ECG utilizing a median value analysis is by far excellent than that based on a single beat analysis.
