RSNA 2016

Abstract Archives of the RSNA, 2016


GU219-SD-MOB4

Clear Cell Renal Cell Carcinoma: Discrimination from Chromophobe RCC and Papillary on MDCT using CAD Derived Peak Lesion Enhancement Relative to Renal Cortex

Monday, Nov. 28 12:45PM - 1:15PM Room: GU/UR Community, Learning Center Station #4



Heidi Coy, Los Angeles, CA (Presenter) Nothing to Disclose
Jonathan R. Young, MD, Los Angeles, CA (Abstract Co-Author) Nothing to Disclose
Michael L. Douek, MD, MBA, Los Angeles, CA (Abstract Co-Author) Nothing to Disclose
Matthew S. Brown, PhD, Los Angeles, CA (Abstract Co-Author) Nothing to Disclose
James Sayre, PhD, Los Angeles, CA (Abstract Co-Author) Nothing to Disclose
Steven S. Raman, MD, Santa Monica, CA (Abstract Co-Author) Nothing to Disclose
PURPOSE

Currently, all solid enhancing renal lesions without macroscopic fat are treated as malignant. Renal cell carcinoma (RCC) subtypes are a heterogeneous group treated by surgery, ablation or active surveillance with a prognosis based on histology, with clear cell RCC (ccRCC) having the highest incidence and metastatic potential. The purpose of our study is to determine if relative enhancement derived from volumetric 3D lesion contour and a Computer Aided Diagnostic (CAD) algorithm to derive lesion enhancement relative to uninvolved renal cortex can discriminate ccRCC from RCC subtypes (chromophobe RCC (chrRCC) and papillary RCC (pRCC)) on four phase MDCT.

METHOD AND MATERIALS

With IRB approval for this HIPAA-compliant retrospective study, we queried our clinical databases to obtain a cohort of histologically proven renal masses with preoperative MDCT with four phases (unenhanced, corticomedullary (CM), nephrographic (NP), and excretory (EX)). The lesion was segmented in each phase. A CAD algorithm selected a 0.5cm region of interest (ROI) of peak lesion enhancement ≤300HU within the 3D lesion contour. A 0.5cm ROI was placed in normal renal cortex. A radiologist approved lesion contours and ROI placement. Relative enhancement (RE) was calculated as: (lesion ROI-cortex ROI)/ (cortex ROI)* 100%). A model was derived using logistical regression with RE of ccRCC, Onc, and fpAML as input. Discrimination was evaluated using receiver operator characteristic (ROC) curves. 

RESULTS

165 patients (69% men, 31% women) with 168 unique renal masses (105 (63%) ccRCC, 18 (11%) chRCC, 45 (27%) pPRCC were analyzed. Mean lesion size in ccRCC= 3.1 cm (range 1.8-6.5), chRCC=2.5 cm (range 0.8-6.0, and pRCC=3.1 cm (range 1.1-6.8). In discriminating ccRCC from chRCC, the model had an AUC of 0.846 (0.735-0.957 95% CI) in the CM phase, 0.827 (0.718-0.937 95% CI) in the NP phase, and 0.848 (0.765-.0.937 95% CI) in the EX phase. In discriminating ccRCC from pRCC, the model had an AUC of 0.958 (0.929-0.986 95% CI) in the CM phase, 0.844 (0.773-0.914 95% CI) in the NP phase, and 0.805 (0.725-.0.884 95% CI) in the EX phase.

CONCLUSION

RE in the CM phase helps discriminate chRCC from ccRCC with an AUC of 0.846 and pRCC from ccRCC with an AUC of 0.958. 

CLINICAL RELEVANCE/APPLICATION

CAD derived RE provides an objective and reproducible measure for the clinician to use when stratifying patients to specific therapeutic pathways, helping to ensure optimal patient outcomes.