Relevance : Patients with chronic endometritis (CE), compared with women without CE, show significantly lower rates of pregnancy (30.8 versus 63.0%) and live births (7.7 versus 51.9%, respectively). Assessing the probability of spontaneous pregnancy in patients with CE after treatment is of scientific and practical interest.

Purpose : To develop a prognostic model for the probability of spontaneous pregnancy in patients with CE and reproductive dysfunction using neural network technology and evaluate its effectiveness.

Material and methods : The work represents a secondary analysis of the results of research “The course and outcomes of pregnancy in patients with CE and reproductive dysfunction who received complex treatment using the drug Superlymph (randomized controlled trial “TULPAN”)”. A total of 875 patients were selected from the electronic database to meet the objectives of this analysis, which was based on the results of a comprehensive examination of the patients. Distribution into comparison groups: I (n = 461, 52.7%) – patients who did not experience spontaneous pregnancy; II (n = 414, 47.3%) – pregnancy occurred spontaneously (within 6 months after treatment).

Results : Pregnancy occurred spontaneously after treatment in 47.3% (414/875), the live birth rate was 39.3% (344/875). Based on the analysis, 12 most significant parameters were identified, which were used to create a model for predicting spontaneous pregnancy based on neural network technology. Of the 414 patients in group II, the prognosis was positive in 390 (94.2%), negative in 24 (5.8%). The prediction accuracy of the developed model was 88.0% (sensitivity – 94.2%, specificity – 82.4%). The informativeness of neural network data analysis was confirmed by ROC analysis – area under the curve (ROC-AUC) = 0.88 (95% confidence interval (CI) = 0.86–0.91; p <0.001). The significant diagnostic role of the oxygenation index according to laser conversion testing data (obtained using the FOTON-BIO spectrometer) is shown – in its absence in the model, the prediction accuracy is reduced to 83.2%. For the purposes of practical application of the model, two online calculators have been developed - for patients and for doctors.

Conclusion : A model for predicting the onset of spontaneous pregnancy in patients with infertility caused by CE using neural network technology has a forecast accuracy of 88% and makes it possible to determine the need for either a repeat course/-s of treatment for CE (if the success rate of spontaneous pregnancy is low) or to make a decision on planning pregnancy (with a high success rate of spontaneous pregnancy).

Anton A. Sukhanov, PhD, is the Head of the Department of Family Planning and Reproduction at the Tyumen Perinatal Center, Tyumen, Russia. He also serves as an Associate Professor in the Department of Obstetrics and Gynecology at Tyumen State Medical University, Ministry of Health of the Russian Federation. With extensive expertise in reproductive medicine, family planning, and obstetrics and gynecology, he is actively involved in clinical practice, medical education, and research, contributing to the advancement of maternal and reproductive healthcare through evidence-based approaches and academic collaboration.