Objective: The purpose of this study was to assess which of a number of methods of measuring attenuation on CT scans is best for prediction of hepatic fat content.
Materials and methods: This retrospective study was approved by our institutional review board. Consecutively registered patients who underwent liver resection for metastatic disease formed the study group. Attenuation measurements were obtained from 12 regions of interest in the liver and three in the spleen on both unenhanced and portal phase contrast-enhanced preoperative hepatic CT images. Hepatic attenuation measurements were analyzed both with and without normalization with the spleen. Normalization included both differences and ratios between hepatic and splenic attenuation values. Pathologic fat content was graded semiquantitatively as a percentage of the nonneoplastic liver parenchyma of the resected specimen. Average attenuation values of the liver were compared with pathologic fat content, as were the differences and ratios between hepatic and splenic attenuation values. Linear regression analysis was conducted on a log-log scale.
Results: Data on 88 patients were analyzed. On unenhanced and contrast-enhanced CT images, all associations between pathologic fat content and attenuation measurements were significant (p < 0.0001). All series of R2 values for unenhanced CT scans were much higher than those for contrast-enhanced CT scans. The R2 values of liver-only measurement were higher than those of hepatic values normalized with splenic values on both unenhanced (0.646-0.649 > 0.523, 0.565) and contrast-enhanced (0.516 > 0.242, 0.344) CT.
Conclusion: Measurement of attenuation of liver only on unenhanced CT scans is best for prediction of pathologic fat content.