Artificial intelligence in in vitro fertilization: Current applications, limitations and future directions

Introduction. Infertility represents a major global health challenge with profound medical, psychological, and social consequences. Despite substantial advances in assisted reproductive techniques, the success rates of in vitro fertilization remain limited, largely due to subjectivity in embryo selection and variability in ovarian stimulation management. Application of artificial intelligence in embryo selection. Artificial intelligence systems enable the analysis of large and complex datasets, including morphological and morphokinetic embryo characteristics derived from static imaging and continuous developmental monitoring. These approaches enhance the standardization of embryo quality assessment and reduce interobserver variability; however, their performance remains constrained by the quality of training data and the incomplete representation of biological determinants of implantation. Role of artificial intelligence in ovarian stimulation management. Algorithms integrating clinical, hormonal, and ultrasonographic parameters facilitate individualized gonadotropin dosing, more accurate determination of final oocyte maturation timing, and reduction of ovarian hyperstimulation risk. Existing studies demonstrate a high level of concordance between algorithm-based recommendations and expert clinical decision-making. Future directions. Further advancements are anticipated through the development of multimodal models integrating morphological, clinical, and genetic data, alongside the implementation of explainable systems to enhance transparency and clinical trust. Conclusion. Currently, artificial intelligence serves primarily as a decision-support tool rather than a substitute for clinical expertise. Its routine clinical implementation requires robust prospective validation and regulatory approval.