Abstract
<jats:p>Background. Non-obstructive azoospermia (NOA) is one of the most complex forms of male infertility, occurring in 10–15% of infertile men. Microdissection testicular sperm extraction (microTESE) is the "gold standard" for obtaining autologous gametes; however, the success rate varies from 40% to 65%, necessitating the search for highly accurate prognostic factors. Purpose. To systematize current approaches to predicting the success of microTESE in patients with NOA. Materials and methods. An analytical review of the literature was conducted with a focus on publications from the last ten years (2016–2026). Studies assessing the prognostic significance of FSH, inhibin B, AMH, their ratio, histological indices (Johnsen score), testicular microenvironment, morphometric nomograms, as well as studies on the application of machine learning algorithms (gradient boosting, Random Forest, neural networks) for predicting microTESE outcomes were analyzed. Results. Isolated hormonal markers were found to have limited prognostic value, whereas the combination of FSH, inhibin B, and AMH, as well as the inhibin B/AMH ratio (cut-off >12.08), demonstrate higher accuracy (sensitivity 72.3%, specificity 83.3%). Histological assessment, including Johnsen score ≥8 and the morphometric nomogram (cut-off >0.489), can increase prognostic specificity to 77.5%. A high density of mast cells in the testicular stroma (>39/mm²) indicates pronounced suppression of spermatogenesis and justifies the transition to sperm donation. Conclusions. A personalized approach integrating hormonal, histological data and AI optimizes patient selection for microTESE, reducing the number of ineffective biopsies; validation of multi-omics biomarkers and the implementation of no-code AI into clinical practice are promising directions.</jats:p>