MobileNet-based Prediction of Preterm Births

Nowadays preterm birth is a serious medical problem since it causes most perinatal deaths and puts people at risk for additional illnesses that could negatively impact their health. The early detection of such cases may prove to be very beneficial in raising society’s overall health standards. In this work, uterine electromyography, also known as Electro Hysterography (EHG), is used to propose such an early diagnosis methodology. A 4
^th
-order band-pass Butterworth filter with a frequency of 0.8 to 5 Hz for each channel is used to filter the raw EHG signals from the EHG channels. Deep Learning (DL) algorithm is used for this purpose. Then, we extract features using DL, and the Quadratic Discriminant validation classifies EHG signals quite reliably. With 100% sensitivity and 99.5% specificity, the suggested framework has a 99.8% accuracy rate for classifying term and preterm birth. So this method provides an early pregnancy observation, preserving the lives of infants, and enhancing mother and baby health. This technology will assist in day-to-day clinical work.