Predicting the risk of decreased ovarian reserve after surgical treatment of patients with deep infiltrative endometriosis using artificial intelligence

Objective. To develop a computer program for predicting the risk of ovarian reserve reduction after surgical treatment of reproductive age patients with deep infiltrative endometriosis (DIE). Material and methods. A case-control study including 60 women of reproductive age with DIE who underwent surgical treatment was performed. Ovarian reserve was evaluated in all patients before surgical treatment and 6 months after it. Group 1 included 30 patients with a statistically significant decrease in ovarian reserve 6 months after surgical treatment for DIE, who were matched with copy-pairs from the cohort of patients without a statistically significant decrease in ovarian reserve, which made up Group 2 (n=30). We performed molecular genetic typing of polymorphic variants of genes encoding proteins involved in the regulation of apoptosis (C-KIT:2600G>A, KITLG:80441 C>T, TP53:Ex4+119 G>C (Arg72Pro) and angiogenesis (VEGF-A:+12143 C>A, VEGF-A:-2578C>A, VEGF-A:-634 G>C, VEGF-A:+936C>T). Results. Using CatBoost machine learning model we selected the most significant clinical and genetic predictors of postoperative reduction of ovarian reserve in patients operated on for DIE: presence of recurrence of ovarian endometrioma; VEGF-A genotype: –634GG; presence of the variant allele -634C of the VEGF-A gene in the genotype; duration of menstruation; TP53:Ex4+119GG genotype; thickness-on M-echo; VEGF-A:+936CC genotype; time since first symptoms; KITLG genotype: 80441CC; history of emergency operative delivery; volume of left and right ovaries according to ultrasound; and blood levels of follicle-stimulating hormone. Then a web application was created using the current stack of web development technologies: HTML, CSS, JavaScript and Python Flask, having an accessible inter-action area. Conclusion. The computer program developed on the basis of artificial intelligence allows to optimize the tactics of management of patients of reproductive age with deep infiltrative endometriosis on the basis of prediction of risks of postoperative decrease of ovarian reserve, as well as to increase the efficiency of the strategy of fertility restoration in patients of this category.