Multiparametric MRI-based radiomics model integrating tumor lesion and periprostatic adipose tissue for predicting bone metastasis in prostate cancer

Purpose To develop and validate a multiparametric MRI (mpMRI)-based radiomics model incorporating features from both the tumor lesion and periprostatic adipose tissue (PPAT) for predicting synchronous bone metastasis (BM) in patients with newly diagnosed prostate cancer (PCa). Methods This retrospective study enrolled 237 patients with histologically confirmed PCa who underwent prostate mpMRI between January 2021 and December 2024. Patients were randomly allocated to a training cohort (n = 165) and a validation cohort (n = 72) in a 7:3 ratio. Univariate and multivariate logistic regression analyses identified independent clinical predictors of BM. Radiomics features were extracted from tumor lesions and PPAT on T2-weighted imaging (T2WI), fat-suppressed T2-weighted imaging (T2WI-FS), and apparent diffusion coefficient (ADC) maps. Separate radiomics models were constructed for intratumoral, PPAT, and combined (intratumoral + PPAT) features. A combined clinical-radiomics model was established by integrating the radiomics score (Rad-score) with independent clinical predictors. Model performance was evaluated using receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA). Results Multivariate analysis identified clinical T stage (OR = 5.00, 95% CI: 1.65–16.70; p = 0.006) and Ki-67 expression (OR = 4.61, 95% CI: 1.63–14.10; p = 0.005) as independent predictors of BM. The intratumoral radiomics model achieved AUCs of 0.928 (training) and 0.835 (validation). The PPAT radiomics model achieved AUCs of 0.859 (training) and 0.842 (validation). The combined radiomics model (intratumoral + PPAT) yielded AUCs of 0.955 (95% CI: 0.927–0.982) and 0.850 (95% CI: 0.745–0.954) in the training and validation cohorts, respectively. The combined clinical-radiomics model demonstrated AUCs of 0.960 (95% CI: 0.934–0.986) and 0.873 (95% CI: 0.785–0.960), respectively. DCA indicated favorable clinical net benefit across a wide range of threshold probabilities. Conclusion The combined model integrating PPAT and tumor radiomics features with clinical predictors demonstrated robust discriminative ability for predicting BM in newly diagnosed PCa. This non-invasive model may provide complementary risk stratification beyond conventional clinical assessment by identifying patients with potentially aggressive disease characteristics and supporting individualized clinical decision-making.