Predicting Lymphatic Uptake Insufficiency in Chronic Bilateral Lower Extremity Edema: A Machine Learning Approach

Objective: Despite the prevalence of bilateral lower extremity lymphedema, no prior studies have utilized machine learning algorithms to predict the development of severe lymphatic insufficiency. This study aimed to predict severe lymphatic insufficiency, defined as the absence of radioactive uptake in the main lymphatic duct on radionuclide lymphoscintigraphy, in patients with bilateral lower extremity edema. Materials and Methods: This retrospective, cross-sectional study included 142 patients with bilateral lower extremity lymphedema and dyspnea on exertion. All participants underwent radionuclide lymphoscintigraphy and myocardial perfusion scintigraphy. Machine learning models—including XGBoost, support vector machines (SVM), multilayer perceptron (MLP), k-nearest neighbors (k-NN), random forests, decision trees, logistic regression, and Naive Bayes—were employed to predict severe lymphatic insufficiency. Model performance was evaluated based on accuracy and ROC-AUC scores. Results: Of the 142 patients (18% male; mean age: 54.67 ± 15.82 years), 72 demonstrated no radioactive uptake in the main lymphatic duct. The machine learning models—XGBoost, SVM, k-NN, logistic regression, and MLP—exhibited high accuracies (1.00, 1.00, 0.976, and 1.00, respectively) and ROC-AUC values (1.00, 1.00, 0.975, and 1.00, respectively). Key predictors of severe lymphatic insufficiency included elevated end-diastolic volume and reduced left ventricular ejection fraction. Conclusions: Machine learning methods successfully predicted severe lymphatic insufficiency in patients with lower extremity lymphedema. The strongest predictors were left ventricular ejection fraction and end-diastolic volume, offering insights for early diagnosis and potential therapeutic interventions.