Related Experiment Video
Updated: Aug 5, 2026
06:03
Integrating Augmented Reality Tools in Breast Cancer Related Lymphedema Prognostication and Diagnosis
Published on: February 6, 2020
Using Machine Learning to Predict Breast Cancer-Related Lymphedema Following Axillary Lymph Node Dissection
Benjamin D Wagner1, Jonlin Chen1, Lillian A Boe2
1Plastic and Reconstructive Surgery Service, Department of Surgery, Memorial Sloan Kettering Cancer Center, New York, NY, USA.
Annals of Surgical Oncology
|July 29, 2026
Summary
Machine learning models significantly outperform traditional methods in predicting breast cancer-related lymphedema (BCRL) after axillary lymph node dissection. This advancement aids in early BCRL identification and risk stratification for improved patient outcomes.
Area of Science:
- Oncology
- Medical Informatics
- Biostatistics
Background:
- Breast cancer-related lymphedema (BCRL) frequently follows axillary lymph node dissection (ALND).
- Existing prediction models for BCRL often lack focus on ALND-specific cohorts or direct comparisons with traditional statistical approaches.
- Accurate prediction of BCRL is crucial for patient management and quality of life.
Purpose of the Study:
- To develop and compare the performance of supervised machine learning (ML) and multivariable logistic regression models for predicting BCRL in patients undergoing ALND.
- To identify key predictors of BCRL development using interpretable ML techniques.
Main Methods:
- Prospective collection of demographic and clinical data from 474 women undergoing unilateral ALND for breast cancer.
- Development and internal validation of supervised ML (random forest) and multivariable logistic regression models.
- Performance evaluation using AUC, accuracy, sensitivity, specificity, and Brier score; Shapley additive explanations for interpretability.
Main Results:
- BCRL developed in 23.8% of patients at a mean of 16.6 months postoperatively.
- The random forest ML model achieved an AUC of 0.83, significantly outperforming logistic regression (AUC = 0.62).
- Key predictors identified by ML included clinical cancer stage, BMI, age, and neoadjuvant chemotherapy.
Conclusions:
- Supervised ML models demonstrate superior predictive performance for BCRL compared to traditional logistic regression in the ALND patient cohort.
- These ML models offer potential for earlier BCRL detection and risk stratification.
- Further research is necessary to enhance predictive accuracy and facilitate clinical integration of ML tools for BCRL management.
