Intrapartum fetal-state classification using long short-term memory neural networks

摘要

We classified perinatal cardiotocography (CTG) records according to their outcome severity using long short-term memory (LSTM) recurrent neural networks. Our classifier classified pathological with 44.7% sensitivity (Se) and 55.3% positive predictive value (PPV), while normal cases had 40.0% Se and 46.7% PPV. The more difficult intermediate MA cases had 33.5% Se and 29.3% PPV.