An application of fuzzy clustering method to cardiotocographic signals classification

Abstract

Cardiotocographic monitoring based on analysis of fetal heart rate, uterine contractions and fetal movements is a primary method for diagnosis of fetal state and prediction of fetal outcome. Visual assessment of signals is very difficult and characterized by intraobserver and interobserver disagreement. In the presented paper, a fuzzy clustering method was applied to cardiotocographic signals classification for fetal outcome prediction. The classifier's fuzzy if-then rules are created based on obtained prototypes. A cross-validation procedure using 100 pairs of learning and testing subsets was applied to validate the results. The obtained results (classification error equal to 21% and sensitivity index equal to 76%) were better in comparison to the Lagrangian SVM method, which is modified version of the best known classification algorithms-Support Vector Machines. © 2011 Springer-Verlag Berlin Heidelberg.