Development and Validation of a Deep Learning System for Staging Liver Fibrosis by Using Contrast Agent-enhanced CT...

Area of Science:

• Radiology
• Artificial Intelligence
• Hepatology

Background:

• Liver fibrosis staging is crucial for patient management.
• Current non-invasive methods have limitations.
• CT imaging offers a readily available modality for liver assessment.

Purpose of the Study:

• To develop and validate a deep learning system (DLS) for liver fibrosis staging using CT images.
• To compare the DLS performance against radiologists and established non-invasive indices.

Main Methods:

• A DLS was trained on CT images from 7461 patients with confirmed liver fibrosis.
• The DLS was validated on a separate dataset of 891 patients.
• Diagnostic performance was assessed using AUROC and Obuchowski indices, comparing DLS to radiologist assessment, APRI, and FIB-4.

Main Results:

• The DLS achieved high accuracy in staging liver fibrosis (79.4%) in test datasets.
• AUROC values for significant fibrosis, advanced fibrosis, and cirrhosis were 0.96, 0.97, and 0.95, respectively.
• The DLS (Obuchowski index, 0.94) significantly outperformed radiologist interpretation and non-invasive markers (range, 0.71–0.81).

Conclusions:

• Deep learning systems can accurately stage liver fibrosis using CT images.
• The developed DLS demonstrates superior performance compared to current methods.
• CT-based DLS shows promise for non-invasive liver fibrosis assessment.