Return To Abstract Listing

Scientific Session 07 — Genitourinary - Prostate

Tuesday, May 2, 2017

Abstracts 2390-3334



3334. Prostate Cancer Volume Correlation on 3-T MultiParametric MRI and Pathology:Using 3D-Printed Molds

Khoshnoodi P*,  Priester A,  Shakeri S,  Asvadi N,  Raman S. University of California, Los Angeles, United States

Address correspondence to P. Khoshnoodi (Pkhoshnoodi@mednet.ucla.edu)

Objective: The purpose of our study was to evaluate correlation of prostate cancer tumor volume of ROIs on prostate multiparametric MRI with tumor volume of concordant foci on whole-mount histopathology and to assess accuracy of MRI in estimating tumor volume stratified by Gleason score and overall suspicion assessment.

Materials and Methods: A HIPAA-compliant, institutional review board–approved study was performed of 114 men who underwent 3-T prostate multiparametric MRI before robotic radical prostatectomy from August 2013 to November 2015. Prostate and suspicious ROIs were segmented on MRI, and overall suspicion assessment was determined on a 1–5 scale using standardized criteria. Tumor volume on multiparametric MRI was calculated by radiologist delineation of ROIs. A 3D patient-specific mold was printed to precisely fit the excised prostate according to MRI prostate segmentation. A pathologist contoured each tumor on all whole-mount histopathology slides. Custom software automatically imported the annotated contoured whole-mount histopathology slides, reconstructed the tumors in 3D, and calculated the tumor volume. A radiologist and pathologist reviewed each case to match each MRI ROI to the concordant focus on whole-mount histopathology slides. Pearson correlation coefficients (ρ) were calculated to determine strength of association between volumes of concordant lesion foci on MRI and whole-mount histopathology findings. Accuracy of MRI in estimating tumor volume was calculated by dividing the volume measured on multiparametric MRI by the volume measured on histopathology. Analyses were conducted using statistical software; p values less than 0.05 were considered significant.

Results: A total of 114 patients had 118 prostate cancer foci (94 index tumors) on whole-mount histopathology slides that were matched with MRI ROIs. For the 118 prostate cancer foci, Gleason score was 3 + 3 in 26 (22%), 3 + 4 in 59 (50%), and 4 + 3 in 33 (28%). Overall suspicion assessment was 2 in three foci (3%), 3 in 37 (31%), 4 in 45 (38%), and 5 in 33 (28%). The median tumor volume on histology and multiparametric MRI was 1.12 and 0.41 cm3, respectively. The ρ between tumor volume on histology and multiparametric MRI was 0.55 overall (p < 0.001), 0.27 for Gleason score of 6 (p = 0.18), 0.52 for Gleason score of 3 + 4 (p < 0.001), and 0.57 for Gleason score of 4 + 3 (p < 0.001) tumors. The ρ was 0.55, 0.48, and 0.65 for tumors with an overall suspicion assessment of 3, 4, and 5, respectively. Accuracy was 0.28, 0.43, and 0.55 for overall suspicion assessment of 3, 4, and 5, respectively (p < 0.001).

Conclusion: Tumor volume can be estimated by measurement on 3-T multiparametric MRI in high grade tumors and in tumors with higher overall suspicion assessment score. Multiparametric MRI consistently underestimates tumor size, and its accuracy is important for targeted biopsy or focal therapy.