Technology for predicting abnormal labor

Abstract

BACKGROUND: Abnormal labor is the most common complication of labor. It occurs in 10-15% cases of labor and is an indication for operative delivery in a third of all cases. Until now, there are no effective ways to predict abnormal labor. Meanwhile, the use of high-tech forecasting methods is not available for a wide range of obstetric institutions in the Russian Federation.
 AIM: The aim of this study was to create a technology for predicting abnormal labor, based on generally available methods of laboratory and instrumental research.
 MATERIALS AND METHODS: Based on the data collected in the Regional Clinical Hospital Perinatal Center, Chita, Russia in 2018-2021, the retrospective analysis of 200 cases of labor was carried out. The total sample was divided into four study groups: 100 women with normal labor activity (group 1), 30 women with uterine inertia (group 2), 30 women with incoordinate uterine activity (group 3), and 50 women with excessive uterine activity (group 4). The groups were comparable in terms of age, anthropometric parameters and extragenital pathology. All women on the eve of labor (1-2 days) underwent general clinical and ultrasound examination. Statistical processing of the results was carried out using the IBM SPSS Statistics version 25.0 software.
 RESULTS: The technology for predicting abnormal labor is implemented based on a multilayer perceptron, with the percentage of incorrect predictions being 21.3%. The structure of the trained neural network included nine input neurons: labor parity, gestational age, leukocyte count, erythrocyte sedimentation rate, total protein concentration, amniotic fluid index, biparietal size, as well as fetal head and abdomen circumference.
 CONCLUSIONS: An integrated approach based on generally available laboratory and instrumental research methods, such as complete blood count and biochemical blood test, as well as ultrasound examination, on the eve of labor allows for predicting the abnormal labor development with an accuracy of up to 70%. The use of this technology in clinical practice will help, in the future, not only to prevent abnormal labor, but also to reduce the incidence of adverse obstetric and perinatal outcomes.