Magnetic resonance imaging radiomic features for recurrent prostate cancer following proton radiation therapy–A pilot study

The role of multiparametric MRI (mp-MRI) for postproton radiation evaluation is unclear. In this pilot study, we characterize the mp-MRI features using the Prostate Imaging-Reporting and Data System (PI-RADS) for recurrent prostate cancer (PCa) following proton radiation therapy. After obtaining IRB approval, we identified 163 consecutive cases who underwent MRI-fusion prostate biopsy at our institution from November 2017 to May 2020. This study evaluated patients with prostate cancer (PCa) with biochemical recurrence following proton radiation. Patients were excluded if they had grossly metastatic disease, metal fragments, implanted devices, or with surgically removed prostates. The mpMRI studies were reviewed in depth and scored by 2 fellowship-trained radiologists. Following MRI-fusion biopsy of lesions of interest (LOI), slides were read by fellowship-trained pathologists. We found 14 patients with 16 lesions who met the study inclusion criteria. The median age was 69 years (range 57–79) and median time to biochemical recurrence was 7.3 years (range 3–13). On post-treatment imaging, decreases in prostate size and diffusely decreased T2 signal intensity were observed, making the use of apparent diffusion coefficient (ADC) and early enhancement at dynamic contrast enhanced (DCE) imaging often necessary for diagnosis of disease recurrence. We identified a total of 16 lesions with PIRADS scores of 3 or higher. Of these lesions, there were 5 PIRADS 3 lesions (4/5 (80%) without prostate cancer), 7 PIRADS 4-5 lesions (6 (86%) had high risk Pca), and 4 lesions with unassigned PIRADS scores (100% had high risk cancers). Among the MRI variables, diffusion weighted imaging (DWI) heterogeneity had the strongest association with recurrence of PCa (P < 0.001). Results of our pilot study showed that the PIRADS scoring system in the postproton radiation therapy setting has some correlations with prostate cancer recurrence; However, the clinical value of these findings are unclear. While definitive PIRADS categorization of lesions demonstrated expected frequency of cancer consistent with the scoring system, all unassigned lesions also harbored malignancy suggesting a cautious approach to PIRADS scoring system in postproton radiation setting. The findings from this study may be validated using a larger cohort.