MRI LI-RADS Version 2018: Impact of and Reduction in Ancillary Features

Area of Science:

• Radiology
• Hepatology
• Medical Imaging

Background:

• Liver Imaging Reporting and Data System (LI-RADS) is crucial for standardizing the imaging and reporting of liver lesions.
• Ancillary features in LI-RADS are used to refine categorization but their necessity and impact require further evaluation.

Purpose of the Study:

• To assess the diagnostic impact of LI-RADS ancillary features on MRI.
• To determine if ancillary features can be reduced without affecting LI-RADS accuracy.

Main Methods:

• Analysis of 222 liver observations from 81 patients using MRI data.
• Comparison of LI-RADS categorization accuracy with and without ancillary features.
• Utilized decision tree analysis and machine learning for feature pruning.
• Assessed interobserver agreement using Krippendorff alpha coefficient.

Main Results:

• Ancillary features altered LI-RADS categories in 7 hepatocellular carcinomas (HCCs) and 51 benign observations.
• No significant difference in HCC distribution across LI-RADS categories with or without ancillary features.
• Five ancillary features identified as noncontributory: corona enhancement, nodule-in-nodule, mosaic architecture, blood products, and fat in mass.
• Interobserver agreement was significantly higher without ancillary features.

Conclusions:

• Ancillary features have a limited impact on LI-RADS categorization accuracy.
• Several ancillary features can likely be removed from LI-RADS without compromising diagnostic performance.
• Removing noncontributory ancillary features may enhance interobserver agreement.