Classification of Parameters Extracted from Cardiotocographic Signals for Early Detection of Metabolic Acidemia in Newborns

Abstract

Fetal acidosis is reflected by the values of umbilical cord pH and base deficit (BDecf): normal recordings (pH over 7.2 and BDecf under 8 mmol/l) and abnormal recordings (pH under 7.2 and BDecf over 8 mmol/l). The purpose of this paper is to present the implementation of an automated system for detecting fetal acidosis in cardiotocographic recordings. The method uses spectral analysis of medium (0.07-0.13 Hz) and high (0.13-1 Hz) frequency spectrum. We implement the algorithm for segments of the recordings without signal loss for better classification. We determined the normalized medium and high frequency components and mid to high frequency ratio. The recordings in the database are divided into a control group (100 normal recordings) and a test group (431 normal or abnormal recordings). A t-test with the p value under 0.05 between the two groups is used to classify the test group. The classification is improved by including the presence of late and prolonged decelerations in the classification process, obtaining the final results, which are comparable to the best ones in current literature.

Keywords

• cardiotocographic signals
• fetal heart rate monitoring
• metabolic acidemia detection
• pattern classification
• spectral analysis