Steatohepatitis Detection from Ultrasound Images Using Attenuation and Backscattering Coefficients

Diagnosing liver nonalcoholic steatohepatitis (NASH) using non invasive procedures is challenging because the visual aspects in US imaging between healthy and affected liver are very much alike. Attenuation and backscattering coefficients are employed to determine the behavior of the US beams into the liver tissue. The values of these coefficients generate features for a classifier. We try to determine if is possible to discriminate between healthy and steatohepatitis patients and if is possible to assess the steatohepatitis grade. Support vector machine is used as a classifier and the Cohen’s kappa statistic evaluates the performance of the classifier. Experiments performed on US images acquired from a lot of 51 biopsied patients shown that these coefficients can successfully be used to distinguish between healthy and moderate/severe steatohepatitis. The role of fibrosis stage and fatty is also investigated. Brunt score is used to assess the steatohepatitis grades. Fibrosis stage and steatosis grade are also determined.