Use of Ultrasound Echotexture Analysis to Determine the Presence of Nonalcoholic Fatty Liver Disease

Abstract

Introduction: It would be desirable to identify a non-invasive method for determining whether and how much fat there is in the liver. Ultrasound imaging has been used for this purpose using descriptive measures. This study aims to evaluate whether raw ultrasound radiofrequency data, using echotexture of a region of interest in the liver is a valid, quantitative tool for assessing fatty liver. Methods: We applied an echotexture technique of ultrasound images to fatty livers in an attempt to determine the presence of fat in the liver. Previous work has shown that the echotexture of ultrasound images of fatty livers are different from normal livers. The echotexture can be quantified using the probability distribution of echo intensities obtained from a Nakagami distribution, defined by a parameter, ‘M', which can be used as a surrogate measure of the underlying echotexture. Increased values of M indicate fibrofatty infiltration. In this study, the raw ultrasound radiofrequency data from individuals with no previous diagnosis of fatty liver disease were acquired by a trained sonographer. The images were analyzed offline using custom software to generate an M parameter from a user-selected region of interest in the liver. The same images were also evaluated by a trained radiologist who graded them as “normal”, “mild” or “moderate” fatty liver. Results: Images from 24 sequentially-recruited subjects were used. Of these, 14 were classified by the radiologist as “normal”, 5 were “mild”, and 5 were “moderate”. None had “severe” fatty liver. The M-parameter for the “normal” group was 0.703 ± 0.068; for the “mild” group was 0.798± 0.032; and for the “moderate” group was 0.796± 0.018. The differences between the “normal” group and the combined group with mild or moderate fatty liver were significant (p < 0.00018). Using a threshold of 0.76, normal livers and livers with mild or moderate fatty infiltration could be separated with 100% sensitivity and 85.7% specificity, with an area under the curve of 0.94. Conclusion: Individuals without fatty liver have significantly different ultrasound echotexture of liver than those who do. Preliminary evidence indicates that ultrasound echotexture analysis may provide a quick, noninvasive alternative method to reliably determine the presence of fatty liver. Currently we are not able to distinguish among the levels of severity with significant probability. The results reported here need to be confirmed in a larger patient sample.