Prediction of clinical narrow pelvis using neural network data analysis

The prevalence of clinical narrow pelvis is 1.3-1.7%, which is associated with an increase in the frequency of childbirth with a large fetus, as well as the appearance of "erased" forms of anatomically narrow pelvis. Fetal-pelvic disproportion is one of the most important factors determining the frequency of intranatal fetal injuries, which determines the relevance of this study. The aim of the study was to evaluate the capabilities of neural network data analysis in predicting a clinical narrow pelvis. A retrospective analysis of 184 birth histories for 2018-2021 was carried out on the basis of the perinatal center of the Regional Clinical Hospital. The total sample was divided into 2 study groups: group 1 icluded 135 patients who gave birth through the natural birth canal; Group 2 included 49 patients whose delivery was complicated by the development of clinically narrow pelvis.Statistically significant parameters such as oligohydramnios, macrosomia, abdominal circumference, fundal height, and fetal head circumference were included in the test database, which formed the basis for training the multilayer perceptron. The structure of the trained neural network included 7 input neurons, one hidden layer containing 9 units, and 2 output neurons (Se=1.00, Sp=0.98, AUC=0.99 [95% CI 0.97-1.00 ], p