Ranista Tongdee MD, Presenter: Nothing to Disclose

Trongtum Tongdee, Abstract Co-Author: Nothing to Disclose

Jin Mo Goo MD, Abstract Co-Author: Nothing to Disclose

Kyongtae Tyler Bae, Abstract Co-Author: Nothing to Disclose

Recently, a histogram analysis method was reported to be useful for differentiating adenoma from metastasis on contrast enhanced CT. The purpose of this study is to evaluate the effects of CT slice thickness and image reconstruction kernel on measurements of mean attenuation and the percentage of negative pixels in a phantom study.

Seven adrenal adenoma phantoms were constructed by blending minced meat, fat, mayonnaise and agar in variable proportions to simulate adenomas with different lipid contents. The phantoms were imaged with 16-row MDCT using 16x0.75 mm collimation, 300 mAs, 120 kVp, 4 slice thicknesses (1, 3, 5, 7 mm), and 3 reconstruction kernels (smooth, medium, sharp). A total of 84 (= 7x4x3) regions of interest were obtained. From the histogram of the attenuation values within each region, the mean attenuation, standard deviations, and the percentage of negative to all pixels (%negative pixel) were computed. Data were plotted and paired t-tests were used to determine if the differences were statistically significant.

All measurements had negative pixels (0.5-35%). The mean attenuation ranged 5-20 HU with 96% (81/84) >10 HU. Increase in %negative pixel was highly correlated (r=0.94, P<0.01) with increase in standard deviations of the attenuation values. But, %negative pixel and mean attenuation values were highly correlated only in certain imaging conditions (images with sharp kernel and 3-7mm thickness). %negative pixel and standard deviations increased together as the reconstruction kernel changed from smooth to sharp and as the slice thickness decreased. In contrast, the mean attenuation was not significantly affected by the kernel and slice thickness.

Our adrenal adenoma phantom study confirmed that the histogram analysis method was more sensitive than the 10 HU mean attenuation method for detecting lipid content. However, the percentage of negative pixels is more variable with the choices of CT imaging parameters than the mean attenuation value. In particular, thinly-sliced images or images reconstructed with sharp kernel may not be appropriate for the histogram method due to increased noise.

Tongdee, R, Tongdee, T, Goo, J, Bae, K, Comparison of CT Histogram Analysis and Mean Attenuation Methods in Characterization of Adrenal Masses: A Phantom Study.  Radiological Society of North America 2004 Scientific Assembly and Annual Meeting, November 28 - December 3, 2004 ,Chicago IL. http://archive.rsna.org/2004/4412135.html