[Fully automatic computer-analysis of electrocardiograms in clinical routine (author's transl)].

A fully automated Ecg recording and interpretation system (Hewlett Packard HP 5) consisting of a central computer and two peripheral units was evaluated during a four-month period. Ecgs were transmitted via public telephone lines, as was the report from central station to peripheral recording and print-out unit. Within a few seconds after recording and print-out unit. Within a few seconds after recording and transmission a preliminary report is available in printed format. The transmitted Ecgs are stored at the central computer and are being retrieved and evaluated manually there. Three cardiologists read the tracings and furnished final reports. 468 Ecgs were thus compared in our test time of 10 days for statistical evaluation with the following results: 1. 73.1% of all statements were read correctly by the computer. 2. There is a large range of diagnostic accuracy between 0% and 100%. 3. The evaluation of Ecg changed by cardiologists reflects the individual experience of the controllers regarding the Ecg statements as well as computer-dependent possibilities like modification etc. of Ecg. 4. As an example the diagnosis "myocardial infarction" is compared with other Ecg computer programs by means of a quality index calculated from sensitivity and specificity: HP 5 with 158.5 points reaches a similar level as the Pipberger program with 162.6 and the Bonner with 159.5 points. 5. Referring to false negative results this program is acceptable in all diagnostic groups with ca. 8%, except conduction defects with 17.7%. 6. Referring to false positive results the diagnostic groups myocardial infarction/ischemia and conduction defects are not satisfactory with 36.5% and 23.4%. 7. In all the tested system seems to be very useful in routine Ecg interpretation of bigger hospitals.