Artificial Intelligence Risk Prediction Tools for Alloplastic Breast Reconstruction

Area of Science:

• Plastic Surgery
• Biomedical Engineering
• Data Science in Healthcare

Background:

• Accurate risk prediction for breast reconstruction with tissue expanders (TEs) is crucial for patient counseling and shared decision-making.
• Alloplastic breast reconstruction involves complex factors influencing patient outcomes.
• Developing robust predictive models can optimize surgical planning and patient management.

Purpose of the Study:

• To develop and evaluate traditional statistical and machine learning (ML) models for predicting complications in alloplastic breast reconstruction.
• To compare the performance of ML models against traditional statistical methods in risk prediction.
• To identify key predictors of complications such as TE loss, infection, and seroma.

Main Methods:

• Retrospective collection of patient data, surgical techniques, and complications for women undergoing immediate TE placement (2017-2023).
• Development of multivariable logistic regression and ML models to predict TE loss, infection, and seroma.
• Optimization of ML models using ten-fold cross-validation and hyperparameter tuning; evaluation via AUC, sensitivity, specificity, and Brier score.

Main Results:

• Analysis of 6,513 immediate TE placements in 4,046 women; complication rates: TE loss (7.6%), infection (10%), seroma (11.5%).
• ML models achieved higher predictive accuracy (AUCs 0.71-0.73) compared to traditional regression (AUCs 0.63-0.69).
• SHAP analysis identified BMI, prepectoral placement, and chemotherapy as significant predictors; models were integrated into nomograms and a web application.

Conclusions:

• Developed accurate risk prediction tools, including nomograms and ML models, for alloplastic breast reconstruction complications.
• Findings support integrating statistical and ML analyses into preoperative assessments for personalized, data-driven patient care.
• Enhanced risk prediction can improve outcomes and patient satisfaction in breast reconstruction surgery.