EHG-Based Preterm Delivery Prediction Algorithm Driven by Transfer Learning

Preterm delivery is currently a global concern of maternal and child health, which directly affects infants’ early morbidity, even death in several severe cases. Therefore, it particularly important to effectively monitor the uterine contraction perinatal pregnant women, make effective prediction timely treatment for possibility preterm delivery. Electromyography (EHG) signal, an measurement predict clinical practice, shows obvious consistency correlation with frequency intensity contraction. This paper proposed deep convolution neural network (DCNN) model based on transfer learning. Specifically, VGGNet model, combined recurrence plot (RP) analysis learning techniques such as “Fine-tune”, marked VGGNet19-I3. Optimized measured term-preterm EHG database, showed good auxiliary performances 78 training test samples, achieved high accuracy 97.00% 100 validation samples.