Multiparametric MRI for the detection of local recurrence of prostate cancer in the setting of biochemical recurrence after low dose rate brachytherapy

PURPOSE Prostate multiparametric magnetic resonance imaging (pMRI) has utility in detecting post-radiotherapy local recurrence. We conducted a multireader study to evaluate the diagnostic performance of mpMRI for local recurrence after low dose rate (LDR) brachytherapy. METHODS A total of 19 patients with biochemical recurrence after LDR brachytherapy underwent 3T endorectal coil mpMRI with T2-weighted imaging, dynamic contrast-enhanced imaging (DCE) and diffusion-weighted imaging (DWI) with pathologic confirmation. Prospective reads by an experienced prostate radiologist were compared with reads from 4 radiologists of varying experience. Readers identified suspicious lesions and rated each MRI detection parameter. MRI-detected lesions were considered true-positive with ipsilateral pathologic confirmation. Inferences for sensitivity, specificity, positive predictive value (PPV), kappa, and index of specific agreement were made with the use of bootstrap resampling. RESULTS Pathologically confirmed recurrence was found in 15 of 19 patients. True positive recurrences identified by mpMRI were frequently located in the transition zone (46.7%) and seminal vesicles (30%). On patient-based analysis, average sensitivity of mpMRI was 88% (standard error [SE], 3.5%). For highly suspicious lesions, specificity of mpMRI was 75% (SE, 16.5%). On lesion-based analysis, the average PPV was 62% (SE, 6.7%) for all lesions and 78.7% (SE, 10.3%) for highly suspicious lesions. The average PPV for lesions invading the seminal vesicles was 88.8% (n=13). The average PPV was 66.6% (SE, 5.8%) for lesions identified with T2-weighted imaging, 64.9% (SE, 7.3%) for DCE, and 70% (SE, 7.3%) for DWI. CONCLUSION This series provides evidence that mpMRI after LDR brachytherapy is feasible with a high patient- based cancer detection rate. Radiologists of varying experience demonstrated moderate agreement in detecting recurrence.