Using Machine Learning as a Seroma Risk Assessment Tool in Prepectoral Breast Reconstruction.

Area of Science:

• Plastic Surgery
• Medical Informatics
• Machine Learning

Background:

• Seroma formation is a common complication following prepectoral breast reconstruction.
• Predicting seroma risk is crucial for optimizing patient care and surgical planning.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting seroma risk after prepectoral breast reconstruction.
• To identify key predictors of seroma formation in this patient population.

Main Methods:

• Retrospective review of institutional data from 318 prepectoral breast reconstructions.
• Development of machine learning models using six algorithms (logistic regression, Naive Bayes, SVM, k-NN, decision tree, random forest) with two feature sets.
• Comparison of model performance using accuracy and area under the receiver operating characteristic curve (AUC).

Main Results:

• The overall seroma rate was 25.58%.
• Significant risk factors included body mass index, mastectomy specimen weight, hypertension, neoadjuvant chemotherapy, and skin-sparing mastectomy.
• A random forest model, utilizing statistically significant variables, achieved the highest AUC of 0.83.

Conclusions:

• Machine learning models can effectively predict patient-specific seroma risk in breast reconstruction.
• These models outperform traditional methods in identifying high-risk patients.
• The developed models facilitate tailored surgical strategies and enhanced follow-up care.