Risk Prediction of Diabetic Foot Amputation Using Machine Learning and Explainable Artificial Intelligence.

Area of Science:

• Medical Informatics
• Computational Medicine
• Diabetes Research

Background:

• Diabetic foot ulcers (DFUs) are a significant cause of lower extremity amputations (LEAs).
• Early identification of high-risk patients is crucial for preventing LEAs.
• Machine learning (ML) and explainable AI offer potential for improved risk prediction.

Purpose of the Study:

• To develop an explainable ML model for predicting LEA risk in DFU patients.
• To identify key predictive factors for LEA in DFU cases.
• To enhance clinical decision-making for early intervention.

Main Methods:

• Retrospective review of 2559 DFU episodes (2012-2017).
• Development of ML models using 51 features for major, minor, and any LEA outcomes.
• Evaluation of model performance using ROC curves, balanced-accuracy, and F1-score; SHapley Additive exPlanations (SHAP) for interpretability.

Main Results:

• ML models achieved high predictive performance (ROC 0.820 for major LEA).
• XGBoost and Gradient Boosted Trees were the top-performing algorithms.
• SHAP analysis identified total white cell count, comorbidity score, and wound characteristics as key predictors.

Conclusions:

• ML models demonstrate strong efficacy in predicting LEA risk for DFU patients.
• Explainable AI enhances clinical utility by highlighting critical risk factors.
• This approach facilitates targeted early interventions to prevent amputations.