Multi-parametric MRI-based Radiomics Predicts the Risk of Recurrent Lower Urinary Tract Obstruction after Benign Prostatic Hyperplasia Surgery

Objective To explore the effectiveness of multi-parameter magnetic resonance imaging (MRI) radiomics in forecasting the likelihood of recurrence after surgery in benign prostatic hyperplasia (BPH). Material and Methods This retrospective study enrolled 134 pathologically confirmed BPH patients from Gansu Provincial People's Hospital (2018–2021), divided into training and validation sets. ROIs were delineated on T2-WI, DWI, and ADC sequences to extract 312 radiomic features. Dimensionality reduction and feature selection were performed using regularization and LASSO regression LR, SVM, and Naive Bayes classifiers were employed to build models, evaluated by ROC-AUC, calibration curves, and decision curve analysis (DCA) for performance, goodness of fit, and clinical utility. Results 30 radiomics features closely related to recurrence were ultimately selected. The AUC of the LR model was 0.953 in the training set and 0.931 in the validation set. For the SVM model, the AUC was 0.982 in the training set and 0.926 in the validation set. For NaiveBayes model, the AUC was 0.850 in the training set and 0.828 in the validation set. Calibration curves indicated good model fitting, and DCA curves showed significant clinical net benefit. Conclusion Multi-parameter MRI radiomics can predict the risk of postoperative recurrence in benign prostatic hyperplasia.