3361 - Detection and Localization of Unfavorable-Histology Prostate Cancer Using MRI and Whole-Mount Histopathology
Presenter(s)
M. Rojo Domingo1,2, A. Dornisch2, C. C. Conlin3, A. Bagrodia4, D. Do2, S. Do2, M. E. Hahn3, J. Javier-Desloges4, K. Kallis5, C. J. Kane4, J. Kim2, M. A. Liss4, J. Liu2, R. Rakow-Penner3, R. Rupareliya2, A. Salmasi4, Y. Song2,6, A. M. Dale3,7, A. Shabaik8, and T. M. Seibert1,2; 1Department of Bioengineering, University of California San Diego, La Jolla, CA, 2Department of Radiation Medicine and Applied Sciences, University of California San Diego, La Jolla, CA, 3Department of Radiology, University of California San Diego, La Jolla, CA, 4Department of Urology, University of California San Diego, La Jolla, CA, 5Department of Surgery, University of California San Diego, La Jolla, CA, 6Department of Electrical and Computer Engineering, University of California San Diego, La Jolla, CA, 7Department of Multimodal Imaging & Genetics, J. Craig Venter Institute, San Diego, CA, 8Department of Pathology, University of California San Diego, La Jolla, CA
Purpose/Objective(s): The primary aim of early prostate cancer (PC) detection is to prevent metastasis. It is critical to detect localized PC that has high metastatic potential: unfavorable-histology PC (uhPC: grade group [GG]=3 or GG2 with cribriform/intraductal features). We evaluate Prostate Imaging Reporting Data System (PI-RADS) and automated Restriction Spectrum Imaging restriction score (RSIrs) for uhPC detection at the patient level with biopsy as reference, and for uhPC localization at the lesion level with whole-mount histopathology (WMHP) from radical prostatectomy as reference. The hypothesis is that PI-RADS lesions and automated RSIrs lesions are useful to detect and localize uhPC for targeted biopsy and treatment.
Materials/Methods: We evaluated patient-level detection of uhPC in a multi-center cohort of patients who underwent MRI and biopsy, and lesion-level localization in a separate single-center cohort of patients with WMHP. At the patient level, the area under the Receiver Operating Characteristic curve (AUC) for PI-RADS and RSI was calculated to compare discrimination of uhPC, and of GG2-unfavorable in particular. At the lesion-level, PI-RADS contours and automatic RSIrs lesions were used to evaluate detection of the most aggressive tumor (index tumor) and all uhPC tumors in WMHP. Sensitivity, positive predictive value (PPV) and the counts of false positive lesions were obtained.
Results: In 1022 patients from 5 centers, patient-level detection of uhPC was similar (p=0.13) for PI-RADS (AUC: 0.73) and RSI (AUC: 0.70), and showed reasonable discrimination in the subset of GG2 uhPC vs. favorable histology PC (fhPC) and benign cases (AUCs: PI-RADS=0.73, RSI=0.74). In 103 patients with WMHP, sensitivity for the index tumor was 86.6% with PI-RADS, 85.4% with RSI, and 92.7% with PI-RADS and RSI combined. 9.7% of PI-RADS lesions were false positives, compared to 19.0% of RSI lesions. For detection of all uhPC tumors, sensitivity was 81.1% with PI-RADS, 85.8% with RSI, and 89.6% with PI-RADS and RSI combined. Among patients with MRI-visible uhPC tumors, 3.9% and 16.9% also had a false positive lesion with PI-RADS and RSI, respectively. MRI-invisible uhPC tumors were typically small (=0.5 mL).
Conclusion: Detection and treatment strategies must focus on uhPC, which represents the class of tumors with substantial metastatic potential. MRI had generally high detection of uhPC, showing the value of MRI as a triage tool that targets biologically aggressive disease. Even for low-grade tumors, MRI discriminated patients with GG2 uhPC (i.e., cribriform/intraductal) from those with fhPC or no cancer. At the lesion level, sensitivity of MRI for uhPC index tumors was 92.7%. Future work to improve the diagnostic pathway of PC will characterize the metastatic potential of MRI-invisible uhPC and improve detection of those that pose a meaningful threat to patient wellbeing.