Source-Newsmedical.net
A groundbreaking study featured in Scientific Reports presents a new artificial intelligence (AI) approach for diagnosing male infertility by analyzing serum hormone levels, offering an innovative alternative to traditional semen analysis. This novel method addresses several limitations of conventional diagnostic procedures, aiming to enhance the accuracy and accessibility of male infertility testing.
Challenges of Conventional Semen Analysis
Infertility impacts approximately 9% of the global population, translating to around 72.4 million individuals. Male infertility contributes to 50% of these cases and is typically diagnosed through semen analysis, which evaluates sperm production, maturation, and overall reproductive health. The World Health Organization (WHO) provides detailed standards for these analyses. However, traditional semen analysis is fraught with challenges, including difficulties in sample collection due to social stigma and the labor-intensive, manual nature of sperm inspection. These issues underscore the need for alternative screening methods.
AI-Based Hormonal Screening
The study analyzed data from 3,662 patients who had undergone both semen analysis and serum hormone assessments. The average age of participants was 36. Among them, 44% had conditions like oligozoospermia or asthenozoospermia, which denote low sperm count or poor sperm motility. Additionally, 12.2% suffered from non-obstructive azoospermia (NOA), and 5.7% had obstructive azoospermia (OA). A notable 46 individuals had cryptozoospermia, characterized by extremely low sperm counts, and only six had ejaculation disorders.
To develop the AI models, the researchers used hormone levels (LH, FSH, PRL, E2) and the testosterone-to-estradiol (T/E2) ratio from the participants. The AI model, Prediction One, achieved an area under the curve (AUC) value of 74.4%, indicating its effectiveness in predicting male infertility. The AutoML Tables model recorded an AUC receiver operating characteristic (AUC ROC) of 74% and an AUC precision-recall (AUC PR) of 77%, demonstrating its accuracy and reliability. Among the hormones, FSH proved to be the most significant predictor, followed by T/E2 and LH. The model successfully identified NOA and mixed hypogonadotropic hypogonadism (MHH) with 100% accuracy and OA with 70% accuracy.
Conclusions and Future Implications
Previous research has explored AI for predicting endocrine conditions, including postoperative outcomes and elevated hormone levels. The current study confirms that AI models can provide precise and consistent predictions for male infertility based on hormonal data. FSH emerged as the most valuable hormone for prediction, aligning with its known role in spermatogenesis, while the T/E2 ratio also proved significant, reflecting its historical use in treating infertility. This AI-based approach not only has the potential to improve infertility diagnostics but may also offer insights into overall male health. Although it is unlikely to replace traditional semen analysis immediately, this method represents a promising alternative that could eventually complement home diagnostic kits.
