Machine learning in semen analysis for male infertility: A systematic review and future directions.

Abstract

Accurate assessment of semen parameters is crucial for diagnosing and managing male infertility. However, conventional manual methods are time-consuming and subject to substantial inter-rater variability. While artificial intelligence (AI) and machine learning (ML) offer promising automation solutions, a comprehensive synthesis of their performance and clinical translation remains limited. This systematic review evaluates the performance, clinical applications, and translational potential of AI- and ML-based methods for semen analysis. Following Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, PubMed, Scopus, and Web of Science were systematically searched for peer-reviewed studies published between 2010 and 2025 evaluating AI/ML methods for semen analysis in male infertility. Data on algorithms, datasets, evaluation metrics, and clinical applications were extracted. Among the 56 included studies, 71% focused on sperm morphology, 13% on motility, 7% on concentration, and 9% addressed multi-parameter sperm assessment. Deep learning (DL), particularly convolutional neural network (CNN)-based models, was used in 71% of the studies. Models for sperm morphology classification frequently achieved accuracies exceeding 90%. Despite these high reported accuracies, the current evidence is largely limited by small, single-center datasets. Current evidence suggests that DL is the most mature AI approach for automated sperm morphology assessment, improving the accuracy, objectivity, and reproducibility of semen analysis. However, broader clinical implementation remains constrained by limited dataset diversity, inconsistent evaluation protocols, insufficient external validation, and poor model generalizability. Future research should prioritize large multicenter datasets, prospective external validation, standardized benchmarking frameworks, and explainable multimodal AI models to facilitate robust clinical translation and implementation in routine reproductive medicine.

Abstract as supplied by the indexing source. © the authors / publisher.