Biomedical subjects
Publications and source records attributed to R Poll.
Common heart arrhythmia monitors are limited to the electrical activity of the ventricles. Severe cardiac malfunctions exhibit typical defects of atrioventricular conduction which often can be efficiently treated with medicine or electrotherapy. The method described here enables the detection of complete atrioventricular blocks and may be used for the improvement of arrhythmia monitoring.
An already published method for automatic processing of a single ECG signal was tested on the base of a data set of a foreign clinic. The yielded results of the P-wave recognition (recognition ratio 87.1% with a failure ratio of 6.0%) confirm fully its reliability reported before. This paper gives a valuation of the method from the clinical point of view by reference to literature considering technical questions. The most important application of this method is the ECG processing under pressure requiring precise time - measuring, classification of separate patterns and recognition even of non-marked parts of the signal course as P-waves, ST-segments and T-waves. Methodical adaption of the method to the special range of application is necessary in all cases. General questions for prospective evaluation of a given automatic recognition system by physicians are discussed briefly.
Algorithms for digital pattern recognition optimized for the demands of the physician are urgently needed. They have to provide high levels of recognition, accuracy, reliability, artefact rejection and flexibility in detecting different types of signal time-course. The microcomputer algorithm presented here works on the principle of Walsh-transformation of signal sections and in-image judging. The algorithm efficiently solves simple tasks, and also recognizes, for instance, ECG P-waves using the same algorithm. A test with 1054 randomly selected outpatient's ECG and with an additional 72 ECGs of inpatients with clinically proved myocardial infarcts produced the following results: The recognition ratio for the R-wave amounted to 98.8% with a failure ratio of 2.3%, while an initial common P-T-pattern was correctly recognized in 80.3% of cases, with a failure ratio of 4.9%. The algorithm was implemented on a Z80 microprocessor and on a single-chip computer Z8.