Developing a nomogram model for predicting non-obstructive azoospermia using machine learning techniques.

Area of Science:

• Reproductive Medicine
• Urology
• Andrology

Background:

• Azoospermia, the absence of sperm, is categorized as obstructive (OA) or non-obstructive (NOA).
• Accurate prediction of NOA is crucial for effective clinical management and patient counseling.
• Biomarker-based predictive models can enhance diagnostic accuracy in azoospermic patients.

Purpose of the Study:

• To develop and validate a predictive model for non-obstructive azoospermia (NOA) using clinical and biochemical markers.
• To identify key predictors of NOA through logistic regression and machine learning analyses.
• To assess the clinical utility and accuracy of a novel nomogram for NOA prediction.

Main Methods:

• Logistic regression and nine machine learning algorithms were employed for predictor identification and model development.
• A dataset of 352 azoospermia patients (152 OA, 200 NOA) was divided into training and validation sets.
• A nomogram model was constructed using identified predictors and validated using ROC curves, calibration plots, and decision curve analysis.

Main Results:

• Follicle-stimulating hormone (FSH) and semen pH were identified as positive predictors of NOA.
• Mean testicular volume (MTV) and inhibin B (INHB) were negatively correlated with NOA.
• The developed nomogram model demonstrated high predictive performance (AUC 0.984 training, 0.976 validation) and clinical utility.

Conclusions:

• A novel nomogram incorporating FSH, INHB, MTV, and semen pH effectively predicts NOA.
• This model provides a valuable, personalized tool for the diagnosis and management of azoospermia.
• The findings support the use of integrated biomarker analysis for improved clinical decision-making in azoospermia.