Machine Learning Algorithms Combining Slope Deceleration and Fetal Heart Rate Features to Predict Acidemia

Electronic fetal monitoring (EFM) is widely used in intrapartum care as the standard method for well-being. Our objective was to employ machine learning algorithms predict acidemia by analyzing specific features extracted from heart signal within a 30 min window, with focus on last deceleration occurring closest delivery. To achieve this, we conducted case–control study involving 502 infants born at Miguel Servet University Hospital Spain, maintaining 1:1 ratio between cases and controls. Neonatal defined pH level below 7.10 umbilical arterial blood. We constructed logistic regression, classification trees, random forest, neural network models combining EFM acidemia. Model validation included assessments of discrimination, calibration, clinical utility. findings revealed that forest model achieved highest area under receiver characteristic curve (AUC) 0.971, but regression had best specificity, 0.879, sensitivity 0.95. In terms utility, implementing cutoff point 31% would prevent unnecessary cesarean sections 51% while missing only 5% acidotic cases. By variables recordings, provide practical tool assist avoiding sections.