Angiomyolipoma with minimal fat on MDCT: can counts of negative-attenuation pixels aid diagnosis?

OBJECTIVE The purpose of this study was to determine whether counts of pixels with subzero attenuation on CT scans can aid in the diagnosis of renal angiomyolipoma with minimal fat. MATERIALS AND METHODS Of 33 angiomyolipomas identified among 719 renal masses resected from 702 patients over 4 years, 15 masses in 15 patients were prospectively diagnosed on the basis of the presence of fat at MDCT. The 18 patients with minimal-fat angiomyolipoma and a matched (age, sex, tumor size) cohort of patients with renal cell carcinoma were included in this study. Three radiologists independently counted the number of pixels with attenuation less than -10, -20, and -30 HU. Receiver operating characteristic analysis of the number of pixels at each cutoff was used to calculate sensitivity, specificity, and positive predictive value with the following criteria: 1, more than 10 pixels less than -20 HU; 2, more than 20 pixels less than -20 HU; 3, more than 5 pixels less than -30 HU. RESULTS Using criterion 1, reader A identified six angiomyolipomas; reader B, five; and reader C, two. The combined sensitivity was 24%; specificity, 98%; and positive predictive value, 69%. Using criterion 2, reader A identified three angiomyolipomas; reader B, four; and reader C, two. The combined sensitivity was 17%; specificity, 100%; and positive predictive value, 100%. Using criterion 3, reader A identified four angiomyolipomas; reader B, four; and reader C, two. The combined sensitivity was 18%; specificity, 100%; and positive predictive value, 100%. CONCLUSION CT findings of more than 20 pixels with attenuation less than -20 HU and more than 5 pixels with attenuation less than -30 HU have a positive predictive value of 100% in detection of angiomyolipoma, but most angiomyolipomas with minimal fat cannot be reliably identified on the basis of an absolute pixel count.