Intelligent Antenatal Fetal Monitoring Model Based on Adaptive Neuro-Fuzzy Inference System Through Cardiotocography

Abstract

In non-stress tests (NST) of antenatal fetal monitoring, obstetricians usually interpret the cardiotocography (CTG), not only in line with fetal monitoring guidelines, but also in accordance with the knowledge gained from individual clinical experience, by which prognosis would be uncertain and ambiguous. Moreover, CTG contains considerable uncertainty and fuzziness, for it is a multi-component and nonlinear complex system which has large amount of information-doped noise. Therefore, an intelligent antenatal fetal monitoring model based on adaptive neural network fuzzy inference system (ANFIS) is presented in this paper. Nine important features were extracted from the CTG case data, the number of fuzzy rules was determined by subtractive clustering, after which the fuzzy system was initialized and adjusted through the self-learning mechanism of neural networks. The efficiency of the proposed model was tested on the antenatal CTG dataset from the UCI repository. The experimental results showed that the method outperformed the existing state-of-the-art antenatal fetal monitoring models and had a significant advantage on the gray stage of “suspicious” discrimination. It indicates that the proposed model has promising learning ability and adaptability for the uncertainty of antenatal fetal monitoring.