Scholarlink

Background <p>Polycystic ovary syndrome (PCOS) is a prevalent and heterogeneous endocrine disorder and a leading cause of female infertility. Impaired oocyte maturation in PCOS patients undergoing in vitro fertilization (IVF) adversely affects fertilization, embryo quality, and pregnancy outcomes. Identifying the underlying factors of suboptimal oocyte maturation in patients with PCOS is challenging due to heterogeneous ovarian responses and complex clinical profiles. Machine learning (ML) provides a promising approach to capture complex interactions among clinical parameters, enabling individualized risk assessment of oocyte maturation and the identification of PCOS-specific factors.</p> Methods <p>A total of 759 patients with PCOS and 809 non-PCOS controls undergoing the GnRH-antagonist protocol were included and randomly divided into training and validation sets. Feature selection was performed using the least absolute shrinkage and selection operator (LASSO) regression and the Boruta algorithm. Based on 13 selected features, ten machine learning models were developed. Model performance was evaluated using multiple metrics, including the area under the receiver operating characteristic curve (AUC), accuracy, precision, sensitivity, specificity, negative predictive value (NPV), and F1 score. SHapley Additive exPlanations (SHAP) were used to interpret feature importance and to distinguish PCOS-specific factors.</p> Results <p>Among ten ML models, the XGBoost model showed the best performance (AUC = 0.868). SHAP analysis identified initial timing of gonadotropin (Gn) administration, basal estradiol (E<sub>2</sub>) levels, age, anti-Müllerian hormone (AMH), and infertility duration as key contributors to oocyte maturation rate prediction in PCOS patients. Individual-level SHAP analysis further revealed substantial inter-patient heterogeneity in feature contributions, enabling personalized interpretation. Comparative modeling and SHAP interpretation in a matched non-PCOS control cohort identified distinct PCOS-specific predictors, underscoring the differential clinical relevance of key variables in PCOS versus non-PCOS populations.</p> Conclusion <p>Our ML model effectively identified the risk of suboptimal oocyte maturation in PCOS patients undergoing GnRH-antagonist protocol. The identified key features provide clinically interpretable insights that may inform individualized ovarian stimulation strategies.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An explainable machine learning model for identifying the risk of suboptimal oocyte maturation in patients with polycystic ovary syndrome undergoing GnRH-antagonist protocol

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月31日

查看全文
  • Longying Zhao,
  • Guoqing Fan,
  • Wanxin Wang,
  • Canquan Zhou

摘要

Background

Polycystic ovary syndrome (PCOS) is a prevalent and heterogeneous endocrine disorder and a leading cause of female infertility. Impaired oocyte maturation in PCOS patients undergoing in vitro fertilization (IVF) adversely affects fertilization, embryo quality, and pregnancy outcomes. Identifying the underlying factors of suboptimal oocyte maturation in patients with PCOS is challenging due to heterogeneous ovarian responses and complex clinical profiles. Machine learning (ML) provides a promising approach to capture complex interactions among clinical parameters, enabling individualized risk assessment of oocyte maturation and the identification of PCOS-specific factors.

Methods

A total of 759 patients with PCOS and 809 non-PCOS controls undergoing the GnRH-antagonist protocol were included and randomly divided into training and validation sets. Feature selection was performed using the least absolute shrinkage and selection operator (LASSO) regression and the Boruta algorithm. Based on 13 selected features, ten machine learning models were developed. Model performance was evaluated using multiple metrics, including the area under the receiver operating characteristic curve (AUC), accuracy, precision, sensitivity, specificity, negative predictive value (NPV), and F1 score. SHapley Additive exPlanations (SHAP) were used to interpret feature importance and to distinguish PCOS-specific factors.

Results

Among ten ML models, the XGBoost model showed the best performance (AUC = 0.868). SHAP analysis identified initial timing of gonadotropin (Gn) administration, basal estradiol (E2) levels, age, anti-Müllerian hormone (AMH), and infertility duration as key contributors to oocyte maturation rate prediction in PCOS patients. Individual-level SHAP analysis further revealed substantial inter-patient heterogeneity in feature contributions, enabling personalized interpretation. Comparative modeling and SHAP interpretation in a matched non-PCOS control cohort identified distinct PCOS-specific predictors, underscoring the differential clinical relevance of key variables in PCOS versus non-PCOS populations.

Conclusion

Our ML model effectively identified the risk of suboptimal oocyte maturation in PCOS patients undergoing GnRH-antagonist protocol. The identified key features provide clinically interpretable insights that may inform individualized ovarian stimulation strategies.