Intrapartum electronic fetal heart rate monitoring to predict acidemia at birth with the use of deep learning

EFM is used in the vast majority of US hospital births, but has significant limitations in achieving its intended goal of preventing intrapartum hypoxic-ischemic injury. Novel deep learning techniques can improve complex data processing and pattern recognition in medicine.We sought to apply deep learning approaches to develop and validate a model to predict fetal acidemia from EFM data.The database was created using intrapartum EFM data from 2006-2020 from a large, multi-site academic health system. Data was divided into training and testing sets with equal distribution of acidemic cases. Several different deep learning architectures were explored.The primary outcome was umbilical artery acidemia, investigated at four clinically meaningful thresholds: 7.20, 7.15, 7.10, and 7.05, along with base excess. Receiver operating characteristic (ROC) curves were generated with area under the curve (AUROC) assessed to determine the performance of the models. External validation occurred utilizing a publicly available Czech database of EFM data.A total of 124,777 EFM files were available; 77,132 had <30% missingness in the last 60 minutes of the EFM tracing; 21,041 were matched to a corresponding umbilical cord gas result, 10,182 of which were timestamped within 30 minutes of the last EFM reading and comprised the final dataset. The prevalence of the outcome in the data was 20.9% with pH <7.2, 9.1% <7.15, 3.3% <7.10, and 1.3% <7.05. The best performing model achieved an AUROC of 0.85 at a pH threshold of <7.05. When predicting the joint outcome of both pH <7.05 and base excess <-10 meq/L, it achieved an AUROC of 0.89. When predicting both pH <7.20 and base excess <-10 meq/L, it achieved an AUROC of 0.87. At pH <7.15 and a PPV of 30%, the model achieved a sensitivity of 90% and a specificity of 48%.Application of deep learning methods to intrapartum EFM analysis achieves promising performance in predicting fetal acidemia. This technology could potentially help improve the accuracy and consistency of EFM interpretation.