Assessing HCC Response to Locoregional Therapies: Diagnostic Performance of LI-RADS Nonradiation Treatment Response Algorithm Version 2024 and the Role of Ancillary Features

Take-Away Points

• ■ Major Focus: To evaluate the diagnostic accuracy of the Liver Imaging Reporting and Data System (LI-RADS) Nonradiation Treatment Response Algorithm (TRA) version 2024 (v2024), both with and without ancillary features, and compare its performance with LI-RADS TRA v2017 and modified Response Evaluation Criteria in Solid Tumors (mRECIST) in assessing hepatocellular carcinoma (HCC) response to locoregional therapy (LRT) at MRI, using pathologic findings from subsequent curative surgery as the reference standard. The study specifically evaluates the potential added value of newly introduced features to LI-RADS Nonradiation TRA v2024, including ancillary features (diffusion- and T2-weighted hyperintensity) and simple enhancement characteristics (masslike enhancement of any degree in any phase), in the assessment of HCC response to LRT.

• ■ Key Results: LI-RADS Nonradiation TRA v2024 with ancillary features demonstrated significantly higher sensitivity and accuracy compared to both LI-RADS Nonradiation TRA v2024 without ancillary features and LI-RADS TRA v2017, as well as higher specificity compared to mRECIST in classifying lesions as pathologically viable or nonviable. The inclusion of ancillary features and simplification of enhancement characteristics improved the characterization of viable tumor, reducing the frequency of LR-TR equivocal category and enabling more definitive response classification.

• ■ Impact: Use of LI-RADS Nonradiation TRA v2024 with ancillary features improves the assessment of HCC response to LTR, particularly in identifying viable tumors, and could better inform subsequent treatment decisions.

Accurate assessment of HCC response to LRT is essential for guiding timely retreatment, determining eligibility for curative therapies, and predicting progression-free survival. Several diagnostic algorithms have been developed to standardize and enhance the objectivity of response evaluation. While the generalized enhancement criteria in mRECIST may overdiagnose treatment-related enhancement as viable tumor, the more specific enhancement criteria in the LI-RADS TRA v2017 can lower the sensitivity of detecting viable HCC, risking poor outcomes after curative surgery for patients erroneously classified as having no viable tumor.

This retrospective single-center study analyzed 231 patients (198 men; median age, 56 years) with 306 HCCs who underwent nonradiation-based LRT, either transarterial chemoembolization or ablation, followed by curative surgery (liver resection or liver transplantation) between 2017 and 2022. Pathologic results were used as the reference standard to classify lesions as viable or nonviable. This study is the first to validate and compare LI-RADS Nonradiation TRA v2024 with other response algorithms while examining the value of ancillary features and simplified enhancement characteristics in identifying viable HCC. Two radiologists, with 5 and 10 years of experience, independently evaluated and classified the treated HCCs using LI-RADS Nonradiation TRA v2024 (with and without ancillary features), LI-RADS TRA v2017 (as viable, equivocal, or nonviable), and mRECIST (as viable or nonviable).

In addition to demonstrating substantial interreader agreement, the diagnostic performance of LI-RADS Nonradiation TRA v2024 with ancillary features demonstrated higher sensitivity for detecting viable HCC for both readers compared to performance without ancillary features (85%–87% vs 80%–81%; P < .002) and improved (82%–83% vs 80%; P < .001) accuracy, without significant change in specificity. Additionally, it showed higher sensitivity than LI-RADS TRA v2017 (85%–87% vs 79%; P < .001) and higher specificity than mRECIST (63%–75% vs 40%–54%; P < .002) for viable HCC detection.

The inclusion of ancillary features in LI-RADS Nonradiation TRA v2024, compared to performance without ancillary features, improved the sensitivity for identifying viable HCC, resulting in a reduction in the LR-TR equivocal cases from 9.8% to 5.6% for reader 1 and from 11.8% to 3.3% for reader 2. Among the ancillary features, diffusion restriction demonstrated sensitivity, specificity, and accuracy of 78%, 64%, and 75%, respectively, in detecting pathologically viable tumors. These findings underscore the value of ancillary features in enhancing the precision of viable HCC assessment.

The simplified enhancement criteria in the LI-RADS Nonradiation TRA v2024 algorithm, defined as masslike enhancement (any degree, any phase) in or along the treated area, without requiring washout or similarity to pretreatment enhancement, demonstrated the highest accuracy (80% [95% CI: 75, 85]) and sensitivity (80% [95% CI: 74, 85]) for detecting pathologic viability. This approach outperformed the nodular, masslike, or internal and thick enhancement with arterial phase hyperenhancement and/or washout criteria outlined in LI-RADS TRA v2017.

A broader, simplified enhancement criteria in LI-RADS Nonradiation TRA v2024, combined with ancillary features including hyperintense signal at diffusion- and T2-weighted imaging, address the challenges posed by LRT-induced alterations in the classic diagnostic features of treatment-naive HCC. This approach reduces posttreatment categorization confusion, as evidenced by a decreased frequency of LR-TR equivocal cases, enabling clearer differentiation between lesions requiring retreatment and those suitable for observation.

While these findings highlight the value of incorporating diffusion-weighted imaging into routine MRI protocols for HCC follow-up after LRT, it is important to note that many patients in this study had chronic hepatitis B, with only 52% with pathologically proven cirrhosis. Therefore, the utility of ancillary features, especially diffusion-weighted imaging, may be limited in advanced cirrhosis. Further studies across different etiologies and stages of cirrhosis are needed to validate these findings, with a focus on assessing partial response on imaging to guide optimal timing for retreatment.

Highlighted Article

• Zhou S, Zhou G, Shen Y, et al. LI-RADS Nonradiation Treatment Response Algorithm Version 2024: Diagnostic Performance and Impact of Ancillary Features. AJR Am J Roentgenol 2024. https://doi.org/10.2214/AJR.24.32035. Published online November 13, 2024.