Electrohysterogram based Term and Preterm Delivery Classification System

Premature birth is a large-scale medical issue by being the cause of most of the perinatal deaths and a risk factor for other medical issues affecting the health and life quality of both mother and infant. Early diagnosis of such cases may prove to be very helpful in improving the health quality of society overall. So, in this study, such early diagnostic methodology is proposed by employing the non-invasive technique of uterine electromyography, which is also called electrohysterography (EHG). The raw EHG signals from three recording channels are first filtered using a 4-pole band-pass Butterworth filter of 0.8Hz-4Hz of each channel. Algorithm of Empirical Mode Decomposition (EMD) is applied as a pre-processing approach and for segmentation purposes. Then using nine statistical features, EHG signals are very accurately classified as ones leading to term or preterm labors under the Quadratic Discriminant at 25% holdout validation. The proposed framework is 98.5% accurate for classification of term and preterm deliveries with 98.48% sensitivity and 98.48% specificity. This technique will help obstetricians in their clinical practices by providing an early assessment of pregnancies and for saving the life of infants and improving the infant and maternal health status.