Three-dimensional Treatment-Planning-based Prediction of Seed Migration to Chest after 125I Seed Brachytherapy for Hepatic Malignancy

PurposeTo develop and validate risk models incorporating clinical and/or imaging parameters based on three-dimensional treatment-planning systems (3D-TPS) to predict the occurrence of 125I seed migration and the number of migrated seeds <2/≥2 to the chest after brachytherapy for patients with malignant hepatic tumors.Methods and MaterialsA total of 480 patients diagnosed with malignant liver tumors receiving 125I seed brachytherapy from July 2010 to May 2020 were retrospectively enrolled. Variables included 3D-TPS-based CT parameters, that is, the distance from the seed to the inferior vena cava (DSI), the distance from the seed to the second hepatic portal (DSP) and the angle from the seed to the second hepatic portal (ASP), and patients’ clinical characteristics, that is, the number of seed implantation procedures (NSP), the maximum number of implanted seeds one time (MAX) and laboratory parameters within 1 week before treatment. Two sets of logistic regression models incorporating clinical and/or imaging variables were developed to predict the occurrence of seed migration and the number of migrated seeds <2/≥2. Model performance was assessed by ROC analysis and decision curve analysis.ResultsCompared with the clinical models, the combined model showed a higher discriminative ability for both the prediction of migration occurrence and number of migrated seeds ≥ 2/<2 to the chest (AUC, 0.879 vs. 0.668, p < 0.05; 0.895 vs. 0.701, p < 0.05). The decision curve analysis results indicated higher net benefits of combined models than clinical models. Variables, including DSI, NSP and pretreatment lymphocyte-to-neutrophil ratio, acted as the most important predictors in combined models.ConclusionsThe proposed combined models based on 3D-TPS improved discriminative abilities for predicting 125I seed migration and number of migrated seeds <2/≥2 to the chest after hepatic brachytherapy, being promising to aid clinical decision-making.