Learning-based fatty liver diagnosis method

Abstract

Absorption coefficient (AC) quantification methods in medical ultrasound (US) are effective for objective routine diagnosis of hepatic steatosis and liver disease. In this paper, we present various learning-based methods to quantify the biomechanical property of the tissue, and introduce clinical solutions for diagnosing hepatic steatosis and liver disease through these methods. First, a method for measuring the AC characteristics of the tissue within a user-specified area is presented. Second, a method to extract the AC of liver parenchyma insensitive to the measured ultrasound view is introduced. Third, a method of reconstructing the spatial characteristic of the target region into a two-dimensional image is suggested. The performance of the proposed methods are verified through simulation and clinical tests. These results demonstrate the clinical potential of the proposed learning-based quantification methods as a new standard for diagnosing hepatic steatosis and liver disease.