Predicting Complications in Breast Reconstruction: Development and Prospective Validation of a Machine Learning Model

Area of Science:

• Oncology
• Plastic Surgery
• Machine Learning in Medicine

Background:

• Nipple-areolar complex (NAC) necrosis is a significant complication following nipple-sparing mastectomy (NSM).
• Accurately identifying patients at risk for NAC necrosis is challenging.
• This complication can negatively impact patient outcomes and satisfaction after breast cancer surgery.

Purpose of the Study:

• To develop and validate a predictive model for NAC necrosis.
• To identify key risk factors associated with NAC necrosis.
• To improve patient selection and surgical planning for NSM.

Main Methods:

• A random-forest classification model was trained on retrospective data from patients undergoing NSM and immediate breast reconstruction.
• Preoperative, operative, and postoperative data were collected and analyzed.
• The model was validated prospectively on a separate cohort of patients.

Main Results:

• The predictive model achieved high accuracy (97%) in prospective validation.
• High specificity (98%) and negative predictive value (98%) were observed.
• Implant weight exceeding 400g was identified as the most significant predictor of NAC ischemia.

Conclusions:

• A machine learning model can accurately predict NAC necrosis post-NSM.
• Implant weight is a modifiable risk factor for NAC necrosis.
• Adjusting implant weight may help mitigate the risk of NAC necrosis and improve surgical outcomes.