Is Multiparametric MRI Useful for Differentiating Oncocytomas From Chromophobe Renal Cell Carcinomas?

Genitourinary Imaging Original Research Galmiche et al. Differentiating Oncocytomas From Chromophobe RCCs Genitourinary Imaging Original Research Chloe Galmiche 1 Jean-Christophe Bernhard 2 Mokrane Yacoub 3 Alain Ravaud 4 Nicolas Grenier 1 François Cornelis 1 Galmiche C, Bernhard JC, Yacoub M, Ravaud A, Grenier N, Cornelis F Keywords: DWI, MRI, oncocytomas, renal cell carcinomas, renal neoplasms DOI:10.2214/AJR.16.16832 C. Galmiche and J. C. Bernhard contributed equally to the study. Received May 27, 2016; accepted after revision August 17, 2016. 1 Department of Radiology, Pellegrin Hospital, Pl Amélie Raba Léon, 33076 Bordeaux, France. Address correspondence to F. Cornelis (francois.cornelis@chu-bordeaux.fr). 2 Department of Urology, Pellegrin Hospital, Bordeaux, France. 3 Department of Pathology, Pellegrin Hospital, Bordeaux, France. 4 Department of Oncology, Saint-André Hospital, Bordeaux, France. AJR 2017; 208:343 350 0361 803X/17/2082 343 American Roentgen Ray Society Is Multiparametric MRI Useful for Differentiating Oncocytomas From Chromophobe Renal Cell Carcinomas? OBJECTIVE. The purpose of this study was to retrospectively evaluate the diagnostic accuracy of multiparametric MRI to differentiate oncocytoma from chromophobe renal cell carcinoma (RCC). MATERIALS AND METHODS. In this retrospective study, 26 histologically confirmed oncocytomas and 16 chromophobe RCCs that underwent full MRI examination were identified in 42 patients (25 men and 17 women) over a 6-year period. Demographic data were recorded. Double-echo chemical-shift, dynamic contrast-enhanced T1- and T2-weighted images, and apparent diffusion coefficient (ADC) maps were reviewed independently by two radiologists blinded to pathologic results. Signal-intensity index (SII), tumor-to-spleen signalintensity ratio, ADC ratio, three wash-in indexes, and two washout indexes were calculated and compared using univariate and ROC analyses. Sensitivity and specificity analyses were performed to calculate diagnostic accuracy. RESULTS. All carcinomas and nine oncocytomas were resected; the remaining 17 oncocytomas were biopsied. Patient age (for oncocytomas: mean, 68.2 years; range, 43 84 years; for RCCs: mean, 60.8 years; range, 20 79 years) and tumor size (for oncocytomas: mean, 35.5 mm; range, 12 98 mm; for RCCs: mean, 37.2 mm; range, 9 101 mm) did not differ significantly across groups (p = 0.132 and 0.265, respectively). Good interobserver agreement was observed for all measurements but four. Oncocytomas presented significantly higher ADC (p = 0.002) and faster enhancement (p = 0.007 0.012) but lower SII (p = 0.03) than carcinomas. This combination provided sensitivity of 92.3% (24/26), specificity of 93.8% (15/16), and accuracy of 92.9% (39/42) for the detection of oncocytomas. CONCLUSION. Multiparametric MRI helps to accurately differentiate oncocytomas from chromophobe RCCs with high sensitivity and specificity. C T features such as stellate scar, spoked-wheel pattern, and segmental enhancement inversion have been reported to be more common in oncocytoma than in chromophobe renal cell carcinoma (RCC) [1, 2]. To our knowledge, no morphologic MRI features have been proven reliable for distinguishing renal oncocytomas from chromophobe RCC. Using conventional MRI, Rosenkrantz et al. [3] were unable to accurately differentiate these renal tumor subtypes, suggesting that they exhibit similar features. Even segmental enhancement inversion, which corresponds to a tumor segment with lower signal intensity in the arterial phase and higher signal intensity at the early excretory phase, has been reported to be nonspecific for diagnosis of renal oncocytoma [4]. For tumors with hyperintense central areas on T2-weighted images compatible with a stellate central scar, double-echo chemical-shift MR sequences and analysis of late central enhancement (i.e., segmental inversion) showed a positive predictive value of 95% for oncocytoma [5]. Moreover, some authors have recently proposed to add functional parameters in multiparametric MRI using chemical-shift, DWI, and contrastenhanced sequences to help improve differentiation of renal tumors [5 8]. Cornelis et al. [6] reported that several parameters obtained after contrast injection provided independent discrimination of these tumor subtypes with a high specificity on a limited number of tumors. The purpose of our study was to retrospectively evaluate the diagnostic accuracy of multiparametric MRI to differentiate his- AJR:208, February 2017 343

Galmiche et al. tologically proven oncocytoma from chromophobe RCC. Materials and Methods Patients This single-institution retrospective study was approved by the institutional review board of Pellegrin Hospital. All patient data came from a prospectively maintained database for research on renal cancer (French Research Network on Kidney Cancer and Commission nationale de l informatique et des libertés authorization DR- 203 206), in which patients are included after they provide informed consent. All patients with a diagnosis of oncocytoma or chromophobe RCC imaged with multiparametric MRI at Pellegrin Hospital between November 2009 and December 2015 were included retrospectively for analysis. Patients with multiple tumors or hereditary disorders (e.g., Birt- Hogg-Dubé syndrome) were excluded. MRI Techniques For all patients, MRI was performed in addition to CT because of equivocal findings on CT, such as a stellate scar [2, 3]. All MR examinations were performed on the same 3-T system (Discovery MR750 W, GE Healthcare) with a phased-array body coil (Table 1). T1-weighted sequences were obtained before and after administration of an IV bolus of 0.1 mmol/kg of body weight (0.5 mmol/ml) of gadoteric acid (Dotarem, Guerbet) at a rate of 2 ml/s, followed by a 20-mL saline flush. DWI using multiple b values (b = 0, 100, 400, 800 mm 2 /s) was part of the standard imaging protocol, and pixelwise apparent diffusion coefficient (ADC) maps were generated using a commercially available software-workstation system (DICOM Viewer, Osirix). Qualitative and Quantitative MRI Analysis MR images were independently reviewed by two board-certified radiologists with 1 and 10 years experience, blinded to the pathology results, on a PACS workstation. The reviewers were aware that the tumors could be only oncocytomas or chromophobe RCCs. They first looked at morphologic characteristics of the tumors and recorded size, presence of a central area corresponding to a stellate scar or a central necrosis and its enhancement, presence of segmental inversion whether complete or incomplete, and location of the tumor [5]. For the quantitative analysis, we used the method described by Cornelis et al. [6]. For this purpose, an identical ROI was drawn on tissue components of each tumor, which were assessed on contrast-enhanced T1- and T2-weighted images. ROIs were reported in all sequences and placed on the enhanced portion of the tumor on each phase TABLE 1: MRI Protocol for Renal Tumors Parameter of the contrast-enhanced dynamic T1-weighted sequence. Only components with enhancement greater than 15% between the phases were included in the ROI. Areas of cystic degeneration, necrosis, and hemorrhage were avoided. Mean signalintensity values were recorded in arbitrary units. For all sequences, similar ROIs with similar areas were drawn on the spleen and renal cortex for reference. Signal-intensity ratio on T2-weighted images was calculated using the following equation: Signal-intensity ratio = (SI tumor / SI kidney ) 100, (1) where SI tumor is the signal intensity of the tumor and SI kidney is the signal intensity of the ipsilateral kidney. The intratumoral lipid content was quantitatively measured on an identical ROI matched on in- and opposed-phase images. Signal-intensity index (SII) on T1-weighted images and the tumorto-spleen signal-intensity ratio were calculated using equations 2 and 3 [9]: SII = (TSI in / TSI opposed ) / TSI in 100, (2) where TSI in is the tumor signal intensity on in-phase images and TSI opposed is the tumor signal intensity on opposed-phase images. Tumor-to-spleen signalintensity ratio was calculated as follows: TSR = ([(TSI opposed / SSI opposed ) / (TSI in / SSI in )] 1) 100, (3) MRI Sequences T1-Weighted a T2-Weighted b Dynamic a Diffusion c Plane Axial Axial Axial Axial Fat saturation No No Yes or no Yes TR/TE 5.3/2.4 1.2 12000/81.888 5.3/1.958 5616/69.8 Angle ( ) 12 111 12 90 Slice thickness (mm) 3.4 4 3.4 5 FOV (mm 2 ) 400 360 400 400 Matrix (mm) 224 280 320 384 224 280 128 190 Scan time (s) 20 180 30 180 Delay (s) 0, 40, 120, 250 b Value (s/mm 2 ) 0, 100, 400, 800 a LAVA-Flex enhanced fast gradient-echo 3D sequence (GE Healthcare). b Fast relaxation fast spin-echo imaging. c Echo-planar imaging. where TSR is the tumor-to-spleen signal-intensity ratio, TSI opposed is the tumor signal intensity on opposed-phase images, SSI opposed is the spleen signal intensity on opposed-phase images, TSI in is the tumor signal intensity on in-phase images, and SSI in is the spleen signal intensity on in-phase images. The ADC ratio was calculated to normalize the value using the formula: ADC ratio = (ADC tumor / ADC kidney ) 100, (4) where ADC tumor is the mean ADC of the tumor and ADC kidney is the mean ADC of the ipsilateral kidney. Three contrast-enhanced wash-in indexes were calculated depending on the contrast-enhanced phase using equation 5: Wash-in index = (SI enhanced / SI unenhanced ) / SI unenhanced 100, (5) where SI enhanced is the signal intensity on contrastenhanced images and SI unenhanced is the signal intensity on unenhanced images. An early arterial wash-in index using the contrast-enhanced arterial phase, a parenchymal wash-in index using the parenchymal phase, and a late wash-in index using the excretory phase were obtained. Initial washout index for the parenchymal phase was calculated as follows: Washout index = (SI parenchymal / SI arterial ) / SI arterial 100, (6) where SI parenchymal is the signal intensity in the parenchymal phase and SI arterial is the signal intensity in the arterial phase. A second, later washout index was obtained between the arterial and excretory phases by substituting the variable SI late (signal intensity in the excretory phase) for SI parenchyma in equation 6. 344 AJR:208, February 2017

Pathologic Analysis Differentiating Oncocytomas From Chromophobe RCCs For all tumors, histopathologic diagnosis was available after either a percutaneous biopsy using an 18-gauge automated side-cutting needle (Monopty Disposable Core Biopsy Instrument, Bard) or a resection. A uropathologist reviewed all tissue samples retrospectively. All tumors were fixed in acetic formaldehyde, and tumor fragments were embedded in paraffin after dehydration. Sections measuring 3 4 μm were stained using H and E, followed by immunostaining using vimentin, cytokeratin 7, cluster of differentiation (CD) 10, CD117, human melanoma black 45, melan-a, and Hale colloidal iron stain. Statistical Analysis All demographic (sex, age), clinical, and pathologic patient data issued from the database were correlated with imaging data. Mean values ± SD were evaluated for the distribution of the quantitative continuous variables in each group of patients. Interobserver agreement between the two radiologists was assessed using the intraclass correlation coefficient (ICC) and the 95% Bland-Altman limits of agreement [10, 11]. An ICC above 0.75 was considered to represent good agreement [10]. After the independent interpretation sessions and before calculation of the quantitative parameters, the values obtained by the two radiologists were averaged to obtain a single measurement for each lesion and quantitative parameters were calculated [5]. The statistical significance to compare signal intensity or ADC values and the overall association between each parameter was evaluated using the Kruskal-Wallis test (Stata, StataCorp), and p < 0.05 was considered significant. Generalized logistic regression models were created, incorporating each parameter for multivariate analysis. To assess the diagnostic accuracy of multiparametric analysis for the diagnosis of each tumor subtype, optimal cutoff values for significant parameters were calculated using ROC curve analysis to achieve the highest Youden index [12]. Sensitivity and specificity with 95% CIs for differentiating renal tumors were calculated, as well as accuracy. Results Patients Twenty-six patients with oncocytoma were included, with a mean age of 68.2 years (range, 43 84 years). Among these patients, 65% (17/26) were men and 35% (9/26) were women. Sixteen patients with chromophobe RCC were included, with a mean age of 60.8 years (range, 20 79 years). Among these patients, 50% (8/16) were men and 50% (8/16) women. No significant differences were observed either for age (p = 0.781) or sex (p = 0.324). Tumors The mean tumor size was 35.5 mm (range, 12 98 mm) for oncocytoma and 37.2 mm (range, 9 101 mm) for chromophobe RCCs (p = 0.265) (Figs. 1 and 2). Across both groups, all tumors but two were heterogeneous on T2-weighted images (95.2%, 40/42). A central area was observed in 24 oncocytomas (92.3%, A C E F Fig. 1 79-year-old man with 19-mm oncocytoma in left kidney. Tumor was removed surgically. A, Transverse T2-weighted fast spin-echo MR image shows heterogeneous renal mass (arrow). T2-weighted signal-intensity ratio was measured with ROI on solid tissue (133.47). This ROI was reported on all sequences. B, Apparent diffusion coefficient (ADC) map shows slight restriction of diffusion in this component of renal mass (arrowhead) compared with renal cortex (ADC ratio = 108.77). C and D, Transverse in-phase (C) and opposed-phase (D) MR images show no loss of tumor signal intensity on opposed-phase image (signal intensity index = 84.09, tumor-to-spleen signal-intensity ratio = 3.62). E and F, Transverse fat-suppressed gadolinium-enhanced T1-weighted spoiled gradient-recalled echo MR images in arterial (E) and nephrographic (F) phases show enhancement within mass followed by washout (arterial wash-in index = 38.89, parenchymal wash-in index = 297.22, late wash-in index 3 = 208.33; initial washout index = 186, late washout index = 122). 24/26) and two chromophobe RCCs (12.5%, 2/16). A complete enhancement of this central area was observed in 21 of these 24 oncocytomas (87.5%) and one chromophobe RCC (50%). An incomplete area of enhancement B D AJR:208, February 2017 345

Galmiche et al. TABLE 2: Quantitative Characteristics and Interreader Variability Analysis Oncocytoma (n = 26) Chromophobe RCC (n = 16) Characteristic Reviewer 1 Reviewer 2 ICC 95% CI Reviewer 1 Reviewer 2 ICC 95% CI SI on T2WI Tumor 653 ± 350 (578) 778 ± 650 (597) 0.796 0.537 0.910 625 ± 358 (567) 628 ± 368 (551) 0.981 0.947 0.994 Kidney 601 ± 390 (450) 593 ± 508 (437) 0.795 0.535 0.910 650 ± 445 (482) 627 ± 460 (401) 0.990 0.971 0.997 ADC ( 10 3 mm 2 /s) Tumor 1937 ± 395 (1922) 2140 ± 451 (2094) 0.598 0.033 0.833 1412 ± 317 (1322) 1507 ± 376 (1420) 0.660 0.182 to 0.902 Kidney 2177 ± 280 (2139) 2061 ± 241 (2046) 0.811 0.545 0.922 2014 ± 226 (1978) 2115 ± 375 (2062) 0.635 0.267 to 0.895 Spleen SI In-phase 316 ± 304 (157) 296 ± 251 (174) 0.967 0.923 0.986 497 ± 353 (380) 477 ± 336 (367) 0.996 0.988 0.999 Opposed-phase 337 ± 301 (174) 302 ± 261 (178) 0.981 0.955 0.992 450 ± 309 (439) 474 ± 330 (387) 0.988 0.962 0.996 Tumor SI on T1WI In-phase 300 ± 276 (162) 312 ± 314 (141) 0.959 0.904 0.982 481 ± 370 (297) 438 ± 318 (369) 0.951 0.841 0.985 Opposed-phase 324 ± 276 (175) 274 ± 275 (138) 0.951 0.886 0.979 484 ± 389 (349) 454 ± 339 (383) 0.965 0.886 0.989 Unenhanced 227 ± 207 (131) 216 ± 195 (126) 0.968 0.928 0.986 318 ± 182 (327) 330 ± 239 (276) 0.813 0.442 0.937 Arterial phase 543 ± 460 (345) 495 ± 457 (312) 0.929 0.835 0.969 558 ± 295 (518) 591 ± 386 (500) 0.601 0.143 to 0.860 Parenchymal phase 633 ± 515 (336) 613 ± 524 (363) 0.964 0.916 0.984 704 ± 408 (761) 756 ± 473 (707) 0.828 0.508 0.940 Late phase 571 ± 460 (282) 452 ± 440 (231) 0.944 0.868 0.973 621 ± 353 (589) 578 ± 437 (484) 0.796 0.392 0.931 Note Except where noted otherwise, data are mean values ± SD in arbitrary units. Numbers in parentheses are medians. RCC = renal cell carcinoma, ICC = intraclass coefficient for comparisons between readers 1 and 2, SI = signal intensity, T2WI = T2-weighted imaging, T1WI = T1-weighted imaging, ADC = apparent diffusion coefficient. was observed centrally in the remaining oncocytomas (12.5%, 3/24) and chromophobe RCC (50%, 1/2). The mean time between MRI and pathology was 38 days (3 132 days) for oncocytomas and 58 days (1 221 days) for chromophobe RCCs (p = 0.132). Among the 26 oncocytomas, 15 (57.7%) were biopsied only, and nine (34.6%) were resected without initial biopsy. The remaining two patients (7.7%) underwent both biopsy and surgery because the initial biopsy was nondiagnostic. All chromophobe RCCs were resected and four (25%, 4/16) were also biopsied before surgery. Initially biopsied malignant tumors were removed surgically in all cases with no discrepancies encountered in the final pathologic report. Quantitative Measurements and Interradiologist Agreement Mean ROI size was 93 ± 58 mm 2 for oncocytoma and 153 ± 117 mm 2 for chromophobe RCC (p = 0.144). The ICC was above 0.75 for all measurements but four. Quantitative results for each tumor type are summarized in Table 2. TABLE 3: Quantitative MRI Parameters Quantitative Analysis No significant differences in signal-intensity ratio and tumor-to-spleen signal-intensity ratio were observed between chromophobe RCCs and oncocytomas (p = 0.195 and p = 0.306, respectively). ADC ratios and SII were significantly different between chromophobe RCCs and oncocytomas (p = 0.002 and p = 0.030, respectively). Regarding contrast-enhanced dynamic behavior, contrastenhanced wash-in indexes in the arterial and parenchymal phases were significantly different between the two types of tumors (p = 0.012 and 0.007, respectively), but differences in contrast-enhanced wash-in indexes in the late phase were of borderline significance (p = 0.065). Initial and late washout indexes were not significantly different (p = 0.632 and p = 0.954, respectively). Results are summarized in Table 3 and Figure 3. Diagnostic Accuracy Analyses Optimal cutoff values identified by ROC and corresponding performances for the Parameter Oncocytoma (n = 16) Chromophobe RCC (n = 7) p a T2-weighted signal-intensity ratio 117.4 ± 38.8 (112.7) 104.0 ± 26.6 (98.7) 0.195 ADC ratio 89.7 ± 16.9 (91.5) 70.4 ± 14.4 (74.2) 0.002 Signal intensity index 17.4 ± 26.5 ( 11.1) 3.0 ± 16.1 ( 0.6) 0.030 Tumor-to-spleen signal-intensity ratio 2.3 ± 22.0 ( 0.1) 8.0 ± 27.9 (3.0) 0.306 Wash-in indexes Arterial 158.8 ± 154.8 (92.1) 89.5 ± 53.2 (95.1) 0.012 Parenchymal 203.2 ± 72.8 (215.7) 131.7 ± 66.8 (142.8) 0.007 Late 160.5 ± 74.1 (153.0) 125.8 ± 113.5 (85.5) 0.065 Washout indexes Initial 27.8.8 ± 49.2 (9.6) 24.1 ± 25.8 (19.6) 0.632 Late 11.7 ± 39.5 ( 3.1) 16.1 ± 51.9 ( 7.1) 0.954 Note Data are mean values (in arbitrary units) ± SD. Numbers in parentheses are medians. Statistical significance was set at p < 0.05. RCC = renal cell carcinoma, ADC = apparent diffusion coefficient. a Kruskal-Wallis test. 346 AJR:208, February 2017

Differentiating Oncocytomas From Chromophobe RCCs TABLE 4: Optimal Cutoff Values and Corresponding Performances for Oncocytoma Diagnosis Parameter AUC (95% CI) significant parameters are reported in Table 4. No form of multivariate analysis permitted clear distinction of chromophobe RCC from oncocytoma. The optimal cutoff points for the distinction of oncocytomas from chromophobe RCCs was achieved using a combination of ADC ratio > 88, SII < 1 (or both), and wash-in indexes in the arterial and parenchymal phases. For a correct retrospective diagnosis of oncocytoma, a combination using the identified thresholds of ADC ratio > 88 or SII < 1 plus an arterial wash-in index > 125 or parenchymal wash-in index > 188 was used to correctly identify 24 of 26 oncocytomas and 15 of 16 chromophobe RCCs, giving sensitivity of 92.3% (95% CI, 74.87 99.05%) (24/26) and specificity of 93.8% (95% CI, 76.77 99.84%) (15/16). The positive predictive value was 96% (95% CI, 79.65 99.90) (24/25) and the negative predictive value was 88.2% (95% CI, 63.56 98.54%) (15/17). For the diagnosis of oncocytomas, accuracy was therefore 92.9% (39/42). A combination associating ADC ratio > 88 and SII < 1 plus an arterial washin index > 125 or parenchymal wash-in index > 188, was also tested. This combination provided sensitivity of 53.8% (95% CI, 33.37 73.41%) (14/26), specificity of 100% (95% CI, 79.41 100%) (16/16), positive predictive value of 100% (95% CI, 76.84 100%) (14/14), negative predictive value of 57.1% (95% CI, 37.18 75.54%) (16/28), and accuracy of 71.4% (30/42) for the diagnosis of oncocytomas. For the positive diagnostic of chromophobe RCCs, tumors were first selected by an arterial wash-in index > 125 or parenchymal washin index > 188. Tumors with values above these thresholds were secondarily selected if they had an ADC ratio < 88 or SII > 1. This successive combination provided sensitivity of 93.8% (95% CI, 69.77 99.84%) (15/16), specificity of 80.8% (95% CI, 60.65 93.45%) (21/26), positive predictive value of 75% (95% CI, 50.90 91.34%) (15/20), negative predictive value of 95.5% (95% CI, 77.16 99.88%) Youden Index Cutoff Value Sensitivity (%) Specificity (%) Accuracy (%) ADC ratio 0.803 (0.662 0.944) 88 76.0 84.6 79.0 Signal intensity index 0.707 (0.542 0.871) 1 66.7 68.0 67.5 Arterial wash-in index 0.733 (0.581 0.885) 125 65.4 81.3 71.4 Parenchymal wash-in index 0.752 (0.602 0.903) 188 65.4 87.5 73.8 Note ADC = apparent diffusion coefficient. (21/22), and accuracy of 85.7% (36/42) for the diagnosis of chromophobe RCCs. Discussion Our study demonstrated that multiparametric MRI can be used to confidently identify renal oncocytomas and chromophobe RCCs, overcoming the traditional limitations of overlapping morphologic imaging A features [13]. Peripheral location, completely or predominantly well-circumscribed margins, lack of perinephric fat invasion or renal vein invasion, and central area or segmental inversion after contrast enhancement are all observed in both subtypes [1 3]. In fact, these imaging features are compatible with both the benignity of renal oncocytoma and the indolent behavior of chromophobe RCC. Quantitative parameters derived from multiparametric MRI can be used to propose a correct diagnosis, as has been suggested for various types of renal tumors [6]. The cutoff values of ADC and SII and the faster enhancement identified for oncocytoma showed high specificity. These findings allow the diagnosis of chromophobe RCC to be ruled out when tumors present such features. On the other hand, high sensitivity was observed for tumors presenting slow wash-in indexes, low ADC ratio, or high SII, corre- C D Fig. 2 55-year-old woman with 21-mm chromophobe renal cell carcinoma in right kidney. Tumor was removed surgically. A, Transverse T2-weighted fast spin-echo MR image shows homogeneous renal mass (arrow) with slightly lower signal intensity compared with renal parenchyma (T2-weighted signal-intensity ratio = 96.9). B, ADC map shows restriction of diffusion into renal mass (arrowhead) (ADC ratio = 75.57). C and D, Transverse in-phase (C) and opposed-phase (D) MR images show no significant loss of tumor signal intensity on opposed-phase image (signal intensity index = 12.86, tumor-to-spleen signal-intensity ratio = 10.86). (Fig. 2 continues on next page) B AJR:208, February 2017 347

Galmiche et al. E F Fig. 2 (continued) 55-year-old woman with 21-mm chromophobe renal cell carcinoma in right kidney. Tumor was removed surgically. E and F, Transverse gadolinium-enhanced T1-weighted spoiled gradient-recalled echo MR images in arterial (E) and nephrographic (F) phase (arterial wash-in index = 54.72, parenchymal wash-in index = 122.36, late wash-in index = 76.18, initial washout index = 43.71, late washout index = 13.71) show progressive enhancement and washout. sponding to chromophobe RCC. Using these multiparametric MRI criteria, a noninvasive classification of these tumor subtypes was possible before pathologic results. Such an approach may help to propose surgical resection in the first instance without requiring biopsy, even in cases of equivocal morphologic findings on CT or MRI. In case of oncocytomatosis or Birt-Hogg-Dubé syndrome, in which oncocytomas and chromophobe RCCs may both occur simultaneously [14], multiparametric MRI could be discussed for diagnosis or to target biopsies in the most suspicious lesions. Because tumor growth alone does not appear to be a reliable sign of malignancy, the proposed parameters may be applied to individualize more aggressive tumors for patients under active surveillance [15, 16]. Prospective validation of these findings is needed to confirm the impact of such evaluation in a clinical setting. SII is a quantitative parameter that appears to significantly increase the ability to differentiate oncocytoma from chromophobe RCC. However, a decrease in SI on opposedphase chemical-shift images is not an identifying feature by itself; it is often observed in clear cell RCCs, angiomyolipomas, and papillary RCCs [17]. Our study also found that ADC ratios of oncocytoma are higher than those seen for chromophobe RCC. Such an observation was also reported between oncocytomas and solid RCCs by Taouli et al. [18]. Oncocytomas had significantly higher ADCs compared with solid RCCs (mean ADC, 1.91 ± 0.97 10 3 mm 2 /s for oncocytomas vs 1.54 ± 0.69 10 3 mm 2 /s for solid RCCs; p = 0.0097). Hötker et al. [19] reported that ADC values for clear cell RCC were statistically significantly higher than those for chromophobe, papillary, or unclassified RCC (p < 0.05) but were similar to those for oncocytoma. These results are of limited utility in a clinical setting because of their overlap with other tumors on box-plot analysis, so chemical-shift imaging (with SII calculation) and DWI (including ADC calculation) appear useful only in multiparametric MRI [6]. Although both renal oncocytoma and chromophobe RCC are most often hypovascular relative to the renal cortex at each of the phases of dynamic contrast-enhanced imaging [20], the wash-in of chromophobe RCC appeared significantly lower and slower compared with oncocytoma, as quantified by the arterial and parenchymal wash-in indexes in our study. However, enhancement has been shown to remain significantly higher for both of these types of tumors than for papillary RCC [6, 20]. Only clear cell RCC and low-fat angiomyolipoma present with higher and earlier enhancement, which may complicate identification of these histologies from oncocytoma if a multiparametric reading of MRI is not performed [20 22]. In our study, both renal oncocytoma and chromophobe RCC had heterogeneous signal intensity on T2-weighted images. In some cases, this finding may be related to the presence of a central area, which may be interesting to study. Cornelis et al. [5] reported that all oncocytomas with a central area presented a late central enhancement within the central scar that resulted in a contrast inversion that could be either complete or incomplete. However, all subtypes of RCCs, not only the chromophobe, may also show complete or partial central enhancement. When partial, enhancement was usually seen as a thin rim along the junction between the scar and surrounding tumor tissue. Alone, this imaging feature is not useful because no difference was reported in detection of segmental enhancement inversion between oncocytoma and chromophobe RCC on MRI with a sensitivity and specificity for diagnosis of oncocytoma of 15% and 90%, respectively [4]. However, all carcinomas presenting with a complete inversion of signal intensity within the central area also showed a signal-intensity loss on opposed-phase images, especially after the calculation of SII and tumor-to-spleen signal-intensity ratio; none of the oncocytomas had this characteristic. This observation ruled out oncocytoma when no complete inversion of signal intensity was seen in the central area or when signal intensity decreased on opposed-phase images [5]. For renal masses presenting with a central heterogeneous area compatible with a central scar on contrast-enhanced CT or T2-weighted MR images, Cornelis et al. [5] suggested additional study of the tumor with a delayed (more than 5 min) gadolinium-enhanced T1-weighted and a chemical-shift series to better characterize these lesions. Our study furthers this recommendation, suggesting that DWI can be used to exclude the diagnosis of oncocytoma with high specificity. Our study has some limitations. It was retrospective and may have had selection bias because MRI was performed only if equivocal findings were observed on CT. Given the relatively low frequency of oncocytomas and chromophobe RCC, we included MR images acquired over a period longer than 6 years. The small number of tumors could also limit the conclusions of this study. However, the number of patients included appears consistent with other published series [2 5]. We included large lesions (> 4 cm) although MR characterization is more likely to be used with small renal masses (< 3 4 cm) for which biopsy can be more challenging and where the prevalence of malignancy is lower. Some tumors were only biopsied without further exploration, so a risk of sampling error is present because of the possibility of hybrid tumors involving both chromophobe RCCs and oncocytoma [23]. However, these mixed tumors are rare [24] and tend to be in patients with hereditary disorders (e.g., Birt-Hogg-Dubé), multiple tu- 348 AJR:208, February 2017

Differentiating Oncocytomas From Chromophobe RCCs 250 100 Arbitrary Units Arbitrary Units 200 150 100 50 400 300 200 100 0 SI Ratio Arterial Wash-In Index Parenchymal Wash-In Index ADC Ratio Late Wash-In Index A Arbitrary Units Arbitrary Units 200 150 100 50 mors, or both; no such patient was included in this study. We focused only on the solid tissue component of the tumors to obtain the quantitative parameters. We did not include macroscopic necrosis in our ROI analysis, which probably affects overall enhancement or the mean ADC of the lesion. Our study was conducted at a single institution using only one MRI scanner, which could influence the results when calculating precise cutoff values from MRI-generated signal-intensity values. We used a semiquantitative method for all parameters to normalize the results. As proposed recently, volumetric assessment of the quantitative parameters for the entire tumor was not performed [25]. Because the histologic reference standard in several patients was only results from percutaneous biopsy, misdiagnosis was possible. However, because all pathologic analyses were reviewed by the same uropathologist and disease progression was monitored, the error bias seems to be limited. Moreover, pathologic analysis can be considered as a reference standard because of the systematic use of immunostaining using vimentin, cytokeratin 7, CD10, and CD117 that can exclude chromophobe RCC with a sensitivity of 100% and specificity of 90% [26, 27]. In summary, our study found that multiparametric MRI can provide specific imaging criteria that can be combined to accurately distinguish chromophobe RCC from oncocytoma. Chromophobe RCC can be suspected in a tumor presenting with relatively slow enhancement, low ADC, or both. These findings appear to rule out the diagnosis of oncocytoma with a high specificity. A prospective validation on a larger scale and using different MRI systems is needed before definitive conclusions can be drawn. 50 0 50 100 50 0 SI Index Initial Washout Index TSR Late Washout Index C Fig. 3 Box-and-whisker plots of all quantitative parameters for oncocytomas (dark gray) and chromophobe renal cell carcinomas (light gray). Middle line of box = median, top of box = 75th percentile, bottom of box = 25th percentile, whiskers = 95% CIs, circles = outliers. A, Signal-intensity (SI) ratio of T2-weighted images and apparent diffusion coefficient (ADC) ratio. Differences in ADC ratio were statistically significant. B, SI index and tumor-to-spleen signal-intensity ratio (TSR). Differences in SI index were statistically significant. C, Arterial, parenchymal, and late wash-in indexes. Differences in arterial and parenchymal wash-in indexes were statistically significant. D, Initial and late washout indexes. Acknowledgment We thank Pippa McKelvie-Sebileau for medical editorial services. References 1. Choi JH, Kim JW, Lee JY, et al. Comparison of computed tomography findings between renal oncocytomas and chromophobe renal cell carcinomas. Korean J Urol 2015; 56:695 702 2. Wu J, Zhu Q, Zhu W, Chen W, Wang S. Comparative study of CT appearances in renal oncocytoma and chromophobe renal cell carcinoma. Acta Radiol 2016; 57:500 506 3. Rosenkrantz AB, Hindman N, Fitzgerald EF, Niver BE, Melamed J, Babb JS. MRI features of renal oncocytoma and chromophobe renal cell carcinoma. AJR 2010; 195:[web]W421 W427 4. Schieda N, Al-Subhi M, Flood TA, El-Khodary M, McInnes MD. Diagnostic accuracy of segmental enhancement inversion for the diagnosis of re- B D AJR:208, February 2017 349

Galmiche et al. nal oncocytoma using biphasic computed tomog- in assessing rater reliability. Psychol Bull 1979; 21. Hindman N, Ngo L, Genega EM, et al. Angiomyoli- raphy (CT) and multiphase contrast-enhanced 86:420 428 poma with minimal fat: can it be differentiated from magnetic resonance imaging (MRI). Eur Radiol 12. Youden WJ. Index for rating diagnostic tests. clear cell renal cell carcinoma by using standard 2014; 24:2787 2794 Cancer 1950; 3:32 35 MR techniques? Radiology 2012; 265:468 477 5. Cornelis F, Lasserre AS, Tourdias T, et al. Com- 13. Harmon WJ, King BF, Lieber MM. Renal oncocy- 22. Hakim SW, Schieda N, Hodgdon T, McInnes MD, bined late gadolinium-enhanced and double-echo chemical-shift MRI help to differentiate renal oncocytomas with high central T2 signal intensity from renal cell carcinomas. AJR 2013; 200:830 838 6. Cornelis F, Tricaud E, Lasserre AS, et al. Routinely performed multiparametric magnetic resonance imaging helps to differentiate common subtypes of renal tumours. Eur Radiol 2014; 24:1068 1080 7. Sasiwimonphan K, Takahashi N, Leibovich BC, Carter RE, Atwell TD, Kawashima A. Small (<4 cm) renal mass: differentiation of angiomyolipoma without visible fat from renal cell carcinoma utilizing MR imaging. Radiology 2012; 263:160 168 8. Rosenkrantz AB, Niver BE, Fitzgerald EF, Babb JS, Chandarana H, Melamed J. Utility of the apparent diffusion coefficient for distinguishing clear cell renal cell carcinoma of low and high nuclear grade. AJR 2010; 195:[web]W344 W351 9. Kim JK, Kim SH, Jang YJ, et al. Renal angiomyolipoma with minimal fat: differentiation from other neoplasms at double-echo chemical shift FLASH MR imaging. Radiology 2006; 239:174 180 10. Büsing KA, Kilian AK, Schaible T, Debus A, Weiss C, Neff KW. Reliability and validity of MR image lung volume measurement in fetuses with congenital diaphragmatic hernia and in vitro lung models. Radiology 2008; 246:553 561 11. Shrout PE, Fleiss JL. Intraclass correlations: uses toma: magnetic resonance imaging characteristics. J Urol 1996; 155:863 867 14. Schatz SM, Lieber MM. Update on oncocytoma. Curr Urol Rep 2003; 4:30 35 15. Dodelzon K, Mussi TC, Babb JS, Taneja SS, Rosenkrantz AB. Prediction of growth rate of solid renal masses: utility of MR imaging features preliminary experience. Radiology 2012; 262:884 893 16. Richard PO, Jewett MA, Bhatt JR, Evans AJ, Timilsina N, Finelli A. Active surveillance for renal neoplasms with oncocytic features is safe. J Urol 2016; 195:581 586 17. Karlo CA, Donati OF, Burger IA, et al. MR imaging of renal cortical tumours: qualitative and quantitative chemical shift imaging parameters. Eur Radiol 2013; 23:1738 1744 18. Taouli B, Thakur RK, Mannelli L, et al. Renal lesions: characterization with diffusion-weighted imaging versus contrast-enhanced MR imaging. Radiology 2009; 251:398 407 19. Hötker AM, Mazaheri Y, Wibmer A, et al. Use of DWI in the differentiation of renal cortical tumors. AJR 2016; 206:100 105 20. Vargas HA, Chaim J, Lefkowitz RA, et al. Renal cortical tumors: use of multiphasic contrast-enhanced MR imaging to differentiate benign and malignant histologic subtypes. Radiology 2012; 264:779 788 Dilauro M, Flood TA. Angiomyolipoma (AML) without visible fat: ultrasound, CT and MR imaging features with pathological correlation. Eur Radiol 2016; 26:592 600 23. Bhatnagar A, Rowe SP, Gorin MA, Pomper MG, Fishman EK, Allaf ME. Computed tomography appearance of renal hybrid oncocytic/chromophobe tumors. J Comput Assist Tomogr 2016; 40:513 516 24. Klomp JA, Petillo D, Niemi NM, et al. Birt-Hogg- Dube renal tumors are genetically distinct from other renal neoplasias and are associated with upregulation of mitochondrial gene expression. BMC Med Genomics 2010; 3:59 25. Vargas HA, Delaney HG, Delappe EM, et al. Multiphasic contrast-enhanced MRI: single-slice versus volumetric quantification of tumor enhancement for the assessment of renal clear-cell carcinoma Fuhrman grade. J Magn Reson Imaging 2013; 37:1160 1167 26. Jain S, Roy S, Amin M, et al. Amylase alpha-1a (AMY1A): a novel immunohistochemical marker to differentiate chromophobe renal cell carcinoma from benign oncocytoma. Am J Surg Pathol 2013; 37:1824 1830 27. Ng KL, Morais C, Bernard A, et al. A systematic review and meta-analysis of immunohistochemical biomarkers that differentiate chromophobe renal cell carcinoma from renal oncocytoma. J Clin Pathol 2016; 69:661 671 350 AJR:208, February 2017