Naoki Takahashi MD, Presenter: Nothing to Disclose

Kohei Sasaguri MD, Abstract Co-Author: Nothing to Disclose

Mitsuru Takeuchi MD, PhD, Abstract Co-Author: Nothing to Disclose

Adam Froemming MD, Abstract Co-Author: Nothing to Disclose

Shuai Leng PhD, Abstract Co-Author: Nothing to Disclose

Akira Kawashima MD, Abstract Co-Author: Nothing to Disclose

To evaluate if negative CT attenuation pixel distribution analysis improves detection of fat in small AML.

29 small (<4cm) AML (mean age: 53) and 68 small RCC (46 clear-cell, 22 other, mean age: 59) who underwent unenhanced (NC) and enhanced CT before nephrectomy were included (mean NC-CT slice thickness: 4 mm). CT images were reviewed for presence of macroscopic fat (subjective method). CT pixel distribution analysis was performed using Matlab-based software. First, a largest possible, free-hand ROI was manually placed on the mass on representative NC-CT image. Subsequently, the software systematically generated multiple round overlapping micro-ROIs in the large ROI. Mean HU values and pixel counts under cutoff values in each of multiple micro-ROIs were calculated. Cutoff values tested were 0, -5, -10, -15 and -20 HU. Minimum of the mean HU values and maximum of the pixel counts were recorded (mean-HU method and pixel-count method); these are equivalent to subjectively identifying suspicious areas and placing small ROIs. Size of micro-ROIs were 37 and 49 pixels (28-37mm2) for mean-HU method and 13 and 29 pixels (10-22mm2) for pixel-count method. The mean-HU and pixel-count methods were tested if they can differentiate AML from RCC and/or detect macroscopic fat.

On subjective analysis, 7 AML contained macroscopic fat and 22 did not, while 1 RCC contained macroscopic fat. When AML with macroscopic fat by subjective method were excluded, neither mean-HU or pixel-cont method could differentiate AML from RCC (sensitivity/specificity: 5%/97% or 9%/87%). Mean-HU/pixel-count methods and subject method were concordant for detecting macroscopic fat in all but 1 case (fat in AML only visible on enhanced CT subjectively, and pixel analysis method did not detect fat on NC-CT). Optimal cutoff for macroscopic fat detection were <-15HU and <-10HU for 25 and 49-pixel-ROIs (mean-HU method) and ≥12 pixels of 13-pixel-ROI or ≥19 of 29-pixel-ROI below -10HU, ≥11 of 13 or ≥17 of 29 below -15HU, or ≥11 of 13 or ≥16 of 29 below -20HU (pixel count method).

Negative CT attenuation pixel distribution analysis does not improve detection of fat in small AML beyond subjective method.

Negative CT attenuation value pixel distribution analysis is not useful in differentiating small AML without macroscopic fat from RCC.

Takahashi, N, Sasaguri, K, Takeuchi, M, Froemming, A, Leng, S, Kawashima, A, Limited Utility of Negative CT Attenuation Value Pixel Distribution Analysis Using Unenhanced CT in Diagnosis of Small (<4cm) Angiomyolipoma (AML) without Macroscopic Fat.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14045716.html