MRI-Derived Radiomics Model to Predict the Biochemical Recurrence of Prostate Cancer Following Seed Brachytherapy

Objective: We aimed to investigate the predictive value of imaging features derived from magnetic resonance imaging (MRI) and develop a radiomics model predicting the biochemical recurrence-free survival (BFS) in prostate cancer (PCa) patients treated with seed brachytherapy (seed-BT). Methods: The data of 272 patients with PCa treated with seed-BT at Peking University Third Hospital from 2007 to 2019 was retrospectively investigated. Based on the eligibility criteria, 83 patients were finally included in our study. The cohort was divided into two groups in a ratio of 8:2 (training set: n = 67, test set: n = 16). The Cox survival model combined with the least absolute shrinkage and selection operator (LASSO) algorithm was applied to select the radiomics features from T2WI of pretreatment MRI. A radiomics model with selected features was established to predict the BFS. Results: Nineteen patients experienced biochemical recurrence (BCR) during a median follow-up period of 46 months. Three features with non-zero coefficients were selected from 1598 features and used to construct a radiomics model for BCR prediction. The model accurately predicted the BCR in both the training and test groups, for which the concordance index (C-index) were 0.83 and 0.78, respectively. Receiver operating characteristic (ROC) analysis of the test set was conducted to assess the prediction accuracy. The model achieved a high area under the operator curve (AUC) performance for BCR prediction in the test cohort. Conclusions: Our study revealed the considerable potential of a radiomics model based on MRI-derived imaging features in BCR prediction of PCa patients after seed-BT. Radiomics provides a new perspective to clinicians assessing the outcome of radiotherapy, facilitating accurate prognostic evaluation and preoperative consultation for PCa patients followed by seed-BT.