Abstract
An attempt was made to produce a practical systemfor automatic diagnosis of ECG, whereby operatian for data acquisition and processing was made simple and effective.
The system consists of a three-channel electrocardiograph, a scanner, an analog-to-digital converter (sampling frequency per channel 333 cps), an MADIC-IIA electronic digital computer (storage capacity 4096 words in a magnetic drum) and an X-Y plotter.
The digitalized data of the second heart beat during 4. 2 seconds of recording with Frank's lead system were smoothed by a weighted moving average method. The magnitude changes in spatial ECG vectors was used for automatic pattern recognition of ECG waves. By the application of the above method, the points of beginning and end of the waves were satisfactorily estimated in 97 per cent of 100 patients. Automatic diagnosis was performed on the basis of logical operation employing amplitudes and time intervals of ECG, and 89 per cent of the results of computer diagnosis coincided with those diagnosed by a cardiologist on the conventional lead ECG.
The average time required for data acquisition was 15 minutes and that for pattern recognition and diagnosis was 5 minutes.
Methods for noise elimination and data processing of ECG were discussed.