D23. AI-Driven XGBoost Model for Predicting Lymphedema Risk Post-Immediate Lymphatic Reconstruction

PURPOSE: Lymphedema is a debilitating complication of cancer treatment, significantly impacting patient quality of life and increasing healthcare burden. Current predictive methods lack precision, limiting early intervention opportunities. This study aims to address this gap by developing and validating an artificial intelligence (AI) model using the eXtreme Gradient Boosting (XGBoost) algorithm to predict the occurrence of lymphedema after immediate lymphatic reconstruction (ILR), thereby enhancing clinical decision-making and surgical outcomes. METHODS: We conducted a retrospective analysis of 145 patients who underwent ILR. The dataset was divided into training (70%), validation (20%), and testing (10%) sets to optimize and validate the AI predictive model. The XGBoost algorithm was selected for its efficiency and accuracy with structured data. Model performance was rigorously evaluated using cross-validation on the training and validation sets, followed by a final assessment on the test set. Evaluation metrics included accuracy, sensitivity, and specificity. RESULTS: The AI model demonstrated excellent predictive accuracy, achieving 86.2% accuracy on the test set. The model showed robust performance in identifying patients at risk of developing lymphedema following ILR, indicating its reliability and potential clinical utility. CONCLUSION: We present a highly accurate AI predictive model for assessing lymphedema risk after ILR. The implementation of this XGBoost model in clinical settings could facilitate early identification of high-risk patients, allowing for proactive interventions and personalized patient care. This advancement has the potential to enhance surgical planning, improve patient outcomes, and reduce the healthcare burden associated with lymphedema management.