The effective prediction of preterm labour continues to be a topic for continued research within the area of pregnancy medicine, of which uterine contraction signals have now shown to be insightful towards the inference of a potential preterm. Bio-Magnetomyography(MMG) is a physiological measurement based tool that measures the orthogonal offset of bio-electrical manifestations from uterine contractions and can serve as an insight towards a potential premature delivery. The decoding of the associated physiological signal is an area of substantial research where classical signal processing approaches and metaheuristics optimisation routines have been utilised towards the post-processing and decomposition of the MMG signals. All of which require a degree of expert knowledge and a certain level of tuning and parameter initialisation.

As part of strides towards creating a more automated clinical decision support platform for the predictions of preterm, we employ the use of the Deep Wavelet Scattering(DWS) model, which allows for a deep multiresolution analysis alongside unsupervised feature learning, for the post-processing of candidate MMG signals. The DWS is combined with select pattern recognition-based prediction machines in order to assemble a Clinical Decision Pipeline for the prediction of the states of various pregnancies, with a greater degree of machine intelligence. The patient cohort involved a mixture of patients from a multitude of ethnicities, of whom delivered a mixture of Term and Preterm split within under and over 48 hours of labour imminency. The results show around a 5% increase in the prediction accuracy when compared to the classical methods, in addition to providing a more automated signal processing pipeline for the predictions.