Background: Varicocele is a common, potentially reversible cause of male infertility. However, its impact on semen parameters varies, and conventional diagnostic methods may not always provide an accurate prognosis. Machine Learning (ML) techniques offer the potential to enhance diagnostic precision by identifying complex patterns among clinical and biochemical markers. Objective: This study explores the application of ML in the diagnostic work-up of varicocele, with a particular focus on predicting its presence and associated alterations in semen quality. Additionally, we investigate the role of inflammatory cytokines as potential biomarkers in male infertility. Methods: We analyzed a dataset of 169 subjects (84 with unilateral varicocele and 85 controls), incorporating semen analysis, cytokine profiling, and clinical parameters. Several ML models were trained and compared, including Support Vector Machines (SVM), Random Forest (RF), XGBoost, and a Deep Neural Network (DNN). Two classification tasks were performed: (1) predicting oligoasthenoteratozoospermia (OAT) and (2) detecting varicocele. Model interpretability was enhanced using Local Interpretable Model-Agnostic Explanations (LIME). Results: The DNN achieved the highest performance in predicting varicocele (accuracy = 0.941, precision = 0.967, recall = 0.924), while RF and XGBoost demonstrated balanced precision and recall for OAT prediction (accuracy = 0.98, F1-score = 0.937). Explainability analysis revealed that sperm motility, sperm count, and inflammatory cytokines (IL-6, TNF-α, IL-17A) played significant roles in model predictions, suggesting a potential link between chronic inflammation and spermatogenesis impairment. Conclusion: ML techniques improve the classification of varicocele and OAT, highlighting the relevance of cytokines as additional diagnostic markers. These findings suggest that integrating ML with semen analysis and inflammatory profiling may refine the clinical assessment of male infertility and guide personalized treatment approaches.
Machine learning applications in the diagnostic work-up of varicocele
Cilia N. D.
Writing – Original Draft Preparation
;Salerno V. M.Writing – Review & Editing
;Pallotti F.Writing – Original Draft Preparation
2026-01-01
Abstract
Background: Varicocele is a common, potentially reversible cause of male infertility. However, its impact on semen parameters varies, and conventional diagnostic methods may not always provide an accurate prognosis. Machine Learning (ML) techniques offer the potential to enhance diagnostic precision by identifying complex patterns among clinical and biochemical markers. Objective: This study explores the application of ML in the diagnostic work-up of varicocele, with a particular focus on predicting its presence and associated alterations in semen quality. Additionally, we investigate the role of inflammatory cytokines as potential biomarkers in male infertility. Methods: We analyzed a dataset of 169 subjects (84 with unilateral varicocele and 85 controls), incorporating semen analysis, cytokine profiling, and clinical parameters. Several ML models were trained and compared, including Support Vector Machines (SVM), Random Forest (RF), XGBoost, and a Deep Neural Network (DNN). Two classification tasks were performed: (1) predicting oligoasthenoteratozoospermia (OAT) and (2) detecting varicocele. Model interpretability was enhanced using Local Interpretable Model-Agnostic Explanations (LIME). Results: The DNN achieved the highest performance in predicting varicocele (accuracy = 0.941, precision = 0.967, recall = 0.924), while RF and XGBoost demonstrated balanced precision and recall for OAT prediction (accuracy = 0.98, F1-score = 0.937). Explainability analysis revealed that sperm motility, sperm count, and inflammatory cytokines (IL-6, TNF-α, IL-17A) played significant roles in model predictions, suggesting a potential link between chronic inflammation and spermatogenesis impairment. Conclusion: ML techniques improve the classification of varicocele and OAT, highlighting the relevance of cytokines as additional diagnostic markers. These findings suggest that integrating ML with semen analysis and inflammatory profiling may refine the clinical assessment of male infertility and guide personalized treatment approaches.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.