Ultrasound Image Computerized Analysis for Non-invasive Quantitative Evaluation of Hepatic Fibrosis

Authors:
Publisher website
Google Scholar
Abstract
Background: Assessing the diagnostic value of liver ultrasound image computerized analysis (USICA) for hepatic fibrosis (HF) staging in respect to the “gold standard” provided by liver biopsy (LB). Methods: Two-hundred twenty-eight patients with chronic hepatopathies were prospectively enrolled in the study. All the patients underwent LB and abdominal ultrasound (US). For quantitative US assessment of HF, an image analysis software was developed and 3 parameters were extracted by wavelet processing of the region of interest: mHLlivermHHliver, mHLlivermLLliver, and mHLlivermHLspleen. To assess the relevance of each feature, the support vector machine (SVM) classifiers were employed to discriminate between the 2 severity classes (i.e., incipient F1-F2 vs advanced F3-F4 fibrosis). The statistical significance of the HF staging was assessed using SVM classifiers, in terms of sensitivity (Se), specificity (Sp), and receiver operating characteristic (ROC) curves. Results: A cut-off value of 0.342 of mHLlivermHHliver allowed the discrimination between the incipient and advanced HF with 79.5% Se and 77.4% Sp, at an area under receiver operating characteristic (AUROC) value of 0.867 (P ˂ .001). Conclusion: The proposed USICA using wavelet filter parameters proved to be an innovative method that is useful for the initial noninvasive evaluation and quantification of HF, with the advantages of simplicity, short calculation time, accessibility, and repeatability. The mHLlivermHHliver parameter has demonstrated good accuracy in distinguishing incipient and advanced HF and can be considered an effective non-invasive imaging marker for the assessment of HF in patients with chronic hepatic disease.

This publication has 27 references indexed in Scilit: