Computerized Cardiotocography Analysis System based on Hilbert Transform

The visual analysis of Cardiotocographic (CTG) examinations is a very subjective process. The accurate detection and segmentation of the fetal heart rate (FHR) features and their time correlation with the uterine contractions (UC) allow a better diagnostic and the possibility of anticipation of many problems related to fetal distress. This paper presents a diagnostic aid system based on digital signal processing techniques to detect and segment changes in the FHR and the uterine tone signals automatically. The FHR baseline detection is proposed after pre-processing filtering. The detection line is an auxiliary signal based on the Baseline and a moving average. The Hilbert Transform is then used with adaptive thresholding techniques for identifying fiducial points on the signals. For an antepartum validation database, i.e., exams collected before labor, the positive predictivity value (PPV) found is 96.80% for the FHR decelerations, and 96.18% for the FHR accelerations. For an intrapartum validation database, the PPV found was 91.31% for the uterine contractions, 94.01% for the FHR decelerations, and 100% for the FHR accelerations. For the whole set of exams, PPV and SE were both 100% for the identification of FHR DIP II and prolonged decelerations.