Related Experiment Videos
Development and Internal Validation of an Explainable Machine Learning Model for Predicting Buttock Claudication
Yajing Li1, Hongru Deng2, Yongquan Gu1
1Department of Vascular Surgery, Xuanwu Hospital, Capital Medical University, Beijing 100053, China.
Bioengineering (Basel, Switzerland)
|June 26, 2026
Summary
A new machine learning tool accurately predicts buttock claudication after endovascular aneurysm repair (EVAR). It identifies key risk factors like internal iliac artery embolization, aiding in better surgical planning and patient care.
Area of Science:
- Vascular Surgery
- Medical Informatics
- Machine Learning
Background:
- Buttock claudication is a common complication after endovascular aneurysm repair (EVAR), significantly impacting patient recovery and quality of life.
- Current preoperative risk stratification tools for this complication are limited, hindering individualized patient management.
Purpose of the Study:
- To develop and validate an explainable machine learning framework for predicting postoperative buttock claudication risk in EVAR patients.
- To identify key preoperative predictors of buttock claudication to inform surgical planning and shared decision-making.
Main Methods:
- A retrospective dual-center cohort study of 272 EVAR patients was conducted.
- Data were split into training (70%) and testing (30%) sets, with missing data handled via imputation or complete-case analysis.
- Ten machine learning algorithms were tuned and evaluated using cross-validation, assessing discrimination, calibration, and clinical utility via Decision Curve Analysis (DCA). SHapley Additive exPlanations (SHAP) were used for model interpretability.
Main Results:
- Of 272 patients, 71 (26.1%) developed buttock claudication.
- Independent risk factors identified included iliac artery involvement, male sex, unilateral/bilateral internal iliac artery embolization, and hyperlipidemia. Having >2 distal internal iliac branches was protective.
- The neural network model demonstrated high sensitivity (0.810) and F1 score (0.557), while CatBoost maximized accuracy (0.790) and specificity (0.900). Calibration was acceptable, and DCA indicated clinical utility.
- SHAP analysis confirmed known physiologic risk factors and provided case-level insights.
Conclusions:
- An explainable machine learning framework can accurately stratify the risk of buttock claudication following EVAR.
- Internal iliac artery embolization, iliac involvement, and distal branch anatomy are critical factors influencing this risk.
- The developed tool, accessible via a web calculator, supports perfusion-aware planning and enhances shared decision-making between clinicians and patients.