Bidirectional Long Short-term Memory-based Intelligent Auxiliary Diagnosis of Fetal Health

Abstract

Fetal distress is the main cause of frequent adverse events such as asphyxia and disability. It is particularly important to monitor the fetal intrauterine state in late pregnancy so as to detect fetal abnormalities in time. Cardiotocography (CTG) is an important tool for fetal health assessment in clinical practice. But the accuracy and consistency of CTG are generally not ideal as the correct interpretation is affected by the doctor's clinical experience and ability. Therefore, this paper proposed an intelligent assistant diagnosis algorithm based on fetal heart rate (FHR) signals, which integrated by the input layer construction based on MelFrequency Cepstral Coefficients (MFCCs) and the deep neural network model based on bidirectional long short-term memory (BLSTM). Through 200 sets of intrapartum data from a public database, the results show that the proposed algorithm achieves a high accuracy of 96.25%, which can achieve a pretty well assessment of the fetal state and assist clinicians to make an auxiliary diagnosis of the fetal health.