Scholarlink

Background <p>Chronic post-surgical pelvic pain syndrome (CPSPP) represents a significant clinical challenge affecting a substantial proportion of patients undergoing pelvic surgical procedures. The complex pathophysiology and multifactorial nature of CPSPP necessitate advanced predictive approaches to improve patient outcomes and optimize treatment strategies.</p> Objective <p>This study aimed to develop and validate a machine learning-based prediction model for CPSPP using comprehensive clinical data, with particular emphasis on model interpretability through SHAP (SHapley Additive exPlanations) analysis to identify key risk factors and enhance clinical decision-making.</p> Methods <p>A retrospective cohort study was conducted involving 62 patients who underwent pelvic surgical procedures. Comprehensive clinical data including demographic characteristics, surgical parameters, pain assessments, and postoperative outcomes were collected. Multiple machine learning algorithms were employed to develop predictive models, with performance evaluation using receiver operating characteristic (ROC) analysis. SHAP values were utilized to provide model interpretability and identify the most influential predictive features.</p> Results <p>A total of 62 female patients were analyzed, with a mean age of 47.6 ± 11.8&#xa0;years and BMI of 26.1 ± 4.9&#xa0;kg/m<sup>2</sup>. Following magnetic stimulation therapy, 27.4% achieved significant clinical improvement, with a mean VAS reduction of 2.05 ± 2.03 points. Among the four machine learning models evaluated, XGBoost achieved the highest performance with an AUC of 0.94 in the training set and 0.77 in the validation set. SHAP analysis identified baseline VAS score, stimulation frequency, and intensity as the most influential predictors of treatment response.</p> Conclusions <p>This study presents a novel machine learning approach for predicting CPSPP treatment outcomes with enhanced interpretability through SHAP analysis. The findings contribute to improved understanding of CPSPP risk factors and provide a foundation for personalized treatment strategies and clinical decision support systems.</p> Graphical Abstract <p></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

Machine learning-based prediction model for chronic post-surgical pelvic pain syndrome: a comprehensive analysis using SHAP interpretability

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月27日

查看全文
  • Junhua Xi,
  • Zhen Wang,
  • Zhongle Xu,
  • Yong Shi,
  • Yanbin Zhang

摘要

Background

Chronic post-surgical pelvic pain syndrome (CPSPP) represents a significant clinical challenge affecting a substantial proportion of patients undergoing pelvic surgical procedures. The complex pathophysiology and multifactorial nature of CPSPP necessitate advanced predictive approaches to improve patient outcomes and optimize treatment strategies.

Objective

This study aimed to develop and validate a machine learning-based prediction model for CPSPP using comprehensive clinical data, with particular emphasis on model interpretability through SHAP (SHapley Additive exPlanations) analysis to identify key risk factors and enhance clinical decision-making.

Methods

A retrospective cohort study was conducted involving 62 patients who underwent pelvic surgical procedures. Comprehensive clinical data including demographic characteristics, surgical parameters, pain assessments, and postoperative outcomes were collected. Multiple machine learning algorithms were employed to develop predictive models, with performance evaluation using receiver operating characteristic (ROC) analysis. SHAP values were utilized to provide model interpretability and identify the most influential predictive features.

Results

A total of 62 female patients were analyzed, with a mean age of 47.6 ± 11.8 years and BMI of 26.1 ± 4.9 kg/m2. Following magnetic stimulation therapy, 27.4% achieved significant clinical improvement, with a mean VAS reduction of 2.05 ± 2.03 points. Among the four machine learning models evaluated, XGBoost achieved the highest performance with an AUC of 0.94 in the training set and 0.77 in the validation set. SHAP analysis identified baseline VAS score, stimulation frequency, and intensity as the most influential predictors of treatment response.

Conclusions

This study presents a novel machine learning approach for predicting CPSPP treatment outcomes with enhanced interpretability through SHAP analysis. The findings contribute to improved understanding of CPSPP risk factors and provide a foundation for personalized treatment strategies and clinical decision support systems.

Graphical Abstract