[Automatic computer analysis of fetal cardiac rhythm]

In utero cardiotocographic recording is currently one of the best means of diagnosis of fetal distress. The purpose of automatic tracing of the fetal heart rate is to prevent inter- and intra-individual variability in order to better evaluate the condition of the fetus. This study was carried out with the assistance of a cardiotocograph HP 8040A connected to a microcomputer equipped with a clock card. The heart signal may then be continuously recorded and stored on floppy disks, which are then reread and processed by microcomputer. The data are divided into time intervals (3, 5 or 10 seconds), selected by the physician. At each stage, mean, median and sample variation are calculated. Losses of signal are read. In a second stage, the use of a digital, linear and predictive filter (Kalman filter) applied to heart rates, allows baseline extraction as well as the detection of variations about this baseline (accelerations, decelerations). With the use of this filter on sample variations, it is possible to detect various variability periods, via two different approaches: classification of variability values in three groups: 0-5 bpm, 5-10 bpm, greater than 10 bpm, detection of increases or decreases of the variability as compared to the mean variability during the recording. Comparison of automatic analysis and visual analysis by segmentation is satisfactory in terms of variability. Detection of accelerations and decelerations is more problematic, for two reasons: the notion of baseline, the accuracy of digital detection: 15 bpm, 15 sec. All these problems are discussed.