Diagnostic performance of CT/MRI LI-RADS radiation treatment response algorithm v2024 for Y-90 radioembolization.

OBJECTIVE

To evaluate the performance and inter-reader agreement of the 2024 Liver Imaging Reporting and Data System (LI-RADS) Radiation Treatment Response Algorithm (LR-Radiation-TRA).

MATERIALS AND

METHODS

This was a single-center retrospective cohort study of patients who underwent Y-90 radioembolization as their most recent locoregional therapy prior to liver transplantation. Three abdominal radiologists independently categorized each observation on the most recent post-treatment exams using LR-Radiation-TRA v2024. Diagnostic performance for tumor viability was determined using explant pathology as the reference standard, and inter-reader agreement was computed using intraclass correlation coefficients (ICCs).

RESULTS

124 observations from 98 patients were categorized. Of the 372 total reads (3 per observation), 247 (66.4%, 95% CI: 61.5 - 71.0%) were LR-TR-Nonviable, 95 (25.5%, 95% CI: 21.4 - 30.2%) were -Nonprogressing, 28 (7.5%, 95% CI: 5.3 - 10.7%) were -Viable, and 2 (0.5%, 95% CI: 0.1 - 1.9%) were -Nonevaluable. The average pairwise agreement was 89.8% (85.0 - 92.2%) overall, 88.7% (82.3 - 92.0%) for CT, and 91.2% (83.3 - 94.2%) for MRI. Due to the low frequency of LR-TR-Viable categorization, inter-reader agreement for Viable versus Nonviable/Nonprogressing was low (ICC = 0.269, 95% CI 0.098 - 0.425). Viability on pathology was 31.2%, 42.1%, and 53.6% for Nonviable, Nonprogressing, and Viable respectively (p = 0.02). Excluding Nonprogressing, LR-TR-Viable achieved a sensitivity of 16.3% (95% CI: 10.1 - 25.2%) and specificity of 92.9% (95% CI: 88.2 - 95.8%) for viable tumor. Logistic regression showed higher odds of viability for LR-TR-Viable versus Nonviable lesions (OR = 2.55, 95% CI 1.16 - 5.62, p = 0.02).

CONCLUSION

LI-RADS-Radiation-TRA v2024 has high specificity for tumor viability after Y-90 radioembolization. While absolute agreement is high, inter-reader agreement for viability is low, suggesting the need for clearer definitions and training to potentially improve reliability in treatment response assessment for radiation-treated HCC.