11: Differentiation Clear Cell Renal Cell Carcinoma from Other Common Malignant and Benign Renal Masses on Multiphasic MRI: A Likert Based Multireader Analysis
Location: Poster Hall, Board A3
Width:
Kidney MR score (KMRS) algorithm.
Kidney MR score (KMRS) algorithm for assessing solid renal masses of any size and stage. Each step denotes characteristics that should be investigated in any part of the mass. The most likely diagnoses and likelihood scores are given at the bottom of each column.
Note: A = arterial, AML = angiomyolipoma, ccRCC = clear cell renal cell carcinoma, chrRCC = chromophobe RCC, CM = corticomedullary phase, D = delayed, E = excretory phase, +FS = presence of macroscopic fat, -FS = absence of macroscopic fat, +Opposed = presence of intravoxel fat, -Opposed = absence of intravoxel fat, pRCC = papillary RCC, U = unenhanced phase.
Background: Solid renal masses can be potentially characterized using multiphasic magnetic resonance imaging (MRI). However, qualitative methods remain challenging to standardize. We assessed the diagnostic performance and the interreader agreement of the kidney MR score (KMRS), which combines quantitative and qualitative methods, in differentiating benign and malignant lesions and clear cell renal cell carcinoma (ccRCC) from other common lesions on multiphasic MRI.
Methods: With IRB approval and HIPAA compliance, this observational study comprised non-syndromic patients with 191 pathologically proven and 11 stable, variably fatty renal masses (142 malignant, 54 benign, and 6 uncertain malignant potential). After training on 5 non-study cases, four abdominal radiologists and four abdominal radiology fellows independently interpreted the study cohort and assigned a KMRS to each lesion on a 5-point scale to indicate the probability of benignity, malignancy, and ccRCC, blinded to clinical data. The ancillary findings were used to adjust the final score.
Results: A total of 178 patients (mean age, 62 years ± 17, 116 men) with 202 solid renal masses were included. In distinguishing malignant from benign lesions, a KMRS ≥ 4 had a median sensitivity, specificity, and AUC of 87.5% (83.8-94.6%), 69.5% (64.8-77.8%), and 0.79 (0.77-0.89), respectively. The median sensitivity and specificity of the KMRS ≤ 2 in diagnosing benign lesions were 33.4% (31.5-40.7%) and 95.6% (93.9-99.3%), respectively. A KMRS of 5 had a median sensitivity, specificity, and AUC of 72.3% (65.4-78.3%), 87.0% (84.9-91.6%) and 0.82 (0.77-0.88), respectively for diagnosing ccRCC. The interreader agreement (kappa (κ) score) was 0.59-0.82.
Conclusions: The proposed KMRS had relatively high performance for diagnosing malignant from benign lesions and ccRCC from all other lesions with good to excellent interreader agreement.
.jpg)