Objective Liver Fibrosis Estimation from Shear Wave Elastography.

Area of Science:

• Hepatology
• Medical Imaging
• Artificial Intelligence

Background:

• Non-alcoholic fatty liver disease (NAFLD) is a prevalent cause of diffuse liver disease.
• Liver biopsy is the standard for assessing fibrosis (F0-F4), but it is invasive.
• Noninvasive methods are needed for longitudinal monitoring and risk stratification in NAFLD patients.

Purpose of the Study:

• To develop and evaluate SWE-Assist, an automated framework for non-invasive liver fibrosis staging.
• To improve the accuracy and efficiency of assessing liver stiffness measurements (LSM) using ultrasound shear wave elastography (SWE).
• To classify liver fibrosis stages, specifically identifying stages at or exceeding F2 for clinical decision-making.

Main Methods:

• Development of an automated framework (SWE-Assist) integrating image quality checks, region of interest (ROI) selection, and classification.
• Evaluation of various machine learning classifiers, including random forest, support vector machine, and convolutional neural network (CNN).
• Utilized a dataset of 3,392 images from 328 cases for training and validation.

Main Results:

• The CNN-based approach within SWE-Assist achieved an area under the receiver operating curve (AUROC) of 0.89.
• This significantly outperformed the conventional stiffness-only AUROC of 0.74.
• The automated method requires only one image per decision, compared to ten for the baseline approach.

Conclusions:

• SWE-Assist demonstrates potential for accurate and efficient non-invasive liver fibrosis staging.
• The AI-driven framework shows promise in improving clinical decision-making for NAFLD patients.
• Further validation on larger datasets is recommended to confirm the clinical utility of SWE-Assist.