mTese

Chandra Prasetyo Utomo, Muhamad Fathurrachman, Muhammad Wildan Pratama, Rizqy, Leo Arrasy, Ponco Birowo

SEMAI – mTESE Sperm Retrieval Estimation System is a platform developed to accurately predict the probability of successful surgical sperm retrieval in patients with non-obstructive azoospermia (NOA). NOA accounts for approximately 10–15% of male infertility cases worldwide. Predicting the success of microdissection testicular sperm extraction (mTESE) remains a significant medical challenge due to highly inconsistent clinical predictive indicators, such as testicular volume, hormone levels, and a history of undescended testis (UDT). To address this, the system integrates robust machine learning (ML) models trained on a diverse multicenter dataset, combining internal and external medical data sources from various regions. The acquired predictor variables include age, hormones (FSH, LH, testosterone), testicular volume, varicocele, and UDT. Evaluated through both internal and external validation, these models generate probabilistic outputs that greatly assist in identifying favorable patient profiles. Ultimately, this ML-based risk stratification empowers clinicians to personalize medical decisions, while the utilization of diverse datasets effectively enhances model generalizability and clinical utility.

---