Predicting risk factors associated with preterm delivery using a machine learning model

The evaluation of uterine contraction offers significant information regarding the progression of labour. The occurrence of deliveries before the expected dates leads to undesirable consequences for the mother and fetus. Electrohysterography (EHG) is a non-invasive monitoring scheme generally preferred to detect preterm delivery and reduce the hostile consequences. This paper proposed a novel term and preterm prediction method for predicting preterm deliveries by employing EHG signals. The proposed scheme involves three phases: pre-processing, feature extraction and prediction. Initially, the acquired EHG signals are pre-processed using a band pass filter and wavelet transform to remove noise and artefacts from the EHG signal. The second stage is feature extraction, where the representative features, including Shannon energy, median frequency, time-varying centroid frequency, etc., are extracted. At last, the prediction is carried out using an enhanced sheep flock optimized hybrid extreme artificial neural learning network (ESFHEANL). ESFHEANL is the combination of a hybrid extreme artificial neural learning network (HEANL) and enhanced sheep flock optimization (ESFO) algorithm. It assists in diagnosing the term and preterm birth more accurately. The proposed scheme is implemented in the Python platform and assessed the performance in terms of accuracy, recall, specificity and f-measure using the term-preterm EHG (TPEHG) database. Finally, the experimental outcomes evidenced that the proposed scheme achieved better performance and was employed to diagnose term and preterm births accurately.