A groundbreaking study published in Scientific Reports has unveiled a new artificial intelligence (AI) screening method that accurately predicts male infertility using serum hormone levels, eliminating the need for traditional semen analysis.
Infertility impacts approximately 9% of the global population, translating to around 72.4 million individuals. Male infertility accounts for half of these cases and is typically diagnosed through detailed semen analysis combined with hormonal assays.
Traditional semen analysis, a critical diagnostic tool, assesses sperm production, maturation, and the functionality of seminal pathways and testicular secretory profiles. The World Health Organization’s guidelines set the standards for evaluating semen parameters. However, conventional methods face significant limitations, including the stigma surrounding sample collection and the labor-intensive nature of manual sperm inspection. These factors highlight the urgent need for alternative diagnostic approaches.
Sperm production relies on the intricate balance of testicular and endocrine functions, beginning with the hypothalamo-pituitary-testicular axis. Key hormones involved in spermatogenesis include luteinizing hormone (LH), follicle-stimulating hormone (FSH), prolactin (PRL), testosterone, and estradiol (E2). FSH and LH, released by the anterior pituitary gland in response to gonadotropin-releasing hormone (GnRH), play pivotal roles in stimulating spermatogenesis. Hormones such as inhibin B and testosterone, secreted by Sertoli and Leydig cells respectively, are crucial in this process. Imbalances in these hormones, such as elevated FSH levels without corresponding changes in LH and testosterone, can signal spermatogenesis disorders.
Previous research has established a connection between serum hormone levels and semen analysis profiles. The new study leveraged machine learning (ML) technology to analyze male infertility based solely on hormone levels. The research involved 3,662 patients who had undergone both semen analysis and serum hormone testing. The average age of participants was 36.
The study revealed that 44% of participants had conditions such as oligozoospermia or asthenozoospermia, which indicate low sperm count and poor sperm motility, respectively. Azoospermia, the complete absence of sperm, was present in 12.2% of cases as non-obstructive azoospermia (NOA) and 5.7% as obstructive azoospermia (OA). Additionally, 46 individuals had cryptozoospermia, and six had ejaculation disorders.
The AI models utilized patient age and various hormones—LH, FSH, PRL, E2, and the testosterone-to-E2 ratio (T/E2)—to predict infertility. The AI model based on Prediction One achieved an area under the curve (AUC) value of 74.4%, reflecting a strong performance. The AutoML Tables model reported an AUC receiver operating characteristic (AUC ROC) of 74% and an AUC precision-recall (AUC PR) of 77%, indicating its effectiveness in distinguishing abnormal results.
FSH emerged as the most significant predictor of infertility, followed by T/E2 and LH. The AI models demonstrated a 100% accuracy rate for NOA and MHH predictions in both 2021 and 2022, with a 70% accuracy rate for OA predictions.
The study’s findings suggest that AI can potentially replace traditional semen analysis in some scenarios, providing an accurate, less invasive screening tool. While it is not yet poised to fully replace semen analysis, this AI approach could serve as a valuable alternative, especially for home diagnostic kits. The study underscores the importance of integrating AI in medical diagnostics, not only for predicting infertility but also for assessing overall health.
Related Links:
