Clinical validation of ultrasound backscatter statistics for the assessment of liver fibrosis

Abstract

Introduction Deposition of fibrous tissue in the liver changes with the progression of liver cirrhosis. Ultrasound backscatter imaging may describe the scatterer distribution in the liver, and reflect the severity of liver cirrhosis. However, it is known that the backscatter statistics are often subject to ultrasound imaging settings, region of interest (ROI) for analysis and hepatic steatosis. This study aims to establish a standard imaging procedure and a discriminant model for assessment of liver fibrosis regardless of the degrees of liver fatty changes. Methods 201 patients received elective liver resection were prospectively enrolled in the study. Ultrasound radio-frequency (RF) data of right liver lobe through an intercostal space with the maximum imaging depth of 8cm were collected using the Terason t3000 ™ (convex array) system pre-operatively from each patient. Since significant inter-ROI variation was found when the ROI is less than 10 cm2, all the ROI s were selected such that the ROI area covered as much as possible the liver parenchyma and was larger than 10 cm2. The liver tissue was examined and determined by an experienced pathologist the degrees of liver fibrosis and fatty changes. Among the 201 patients, there were 36 F0, 51 F1, 26 F2, 22 F3 and 66 F4 patients while 62 with mild steatosis (5-30%) and 26 with severe steatosis (> 30%). A FDA-cleared backscatter analysis tool (AmCAD-USTM) was used to calculate the statistics and a stepwise logistic regression model was established to differentiate significant liver fibrosis (≥ F2). Results Three backscatter statistics, including the 95th percentile of the Nakagami (m) statistic, and the 75th and 95th percentiles of the histogram (h) statistics, were selected into the logistic model. Regardless of the liver fatty changes, the Receiver Operating Characteristic (ROC) curve analysis showed a significant discriminant power of the model for evaluation of liver fibrosis - stage 0 versus stage 1-4, AUC 0.77 (p-value Conclusions The liver fibrosis in human results in variation of the scatterer distributions in medical ultrasound imaging. Following a standard imaging procedure, a discriminant model is established and shown to be able to distinguish significantly different degrees of liver fibrosis regardless of the presence of steatosis.