Diagnosis of Hepatocellular Carcinoma Using Gadoxetic Acid-Enhanced MRI Based on LI-RADS Version 2018 and LI-RADS...

The diagnostic performance of the Liver Imaging Reporting and Data System (LI-RADS) version 2018 (v2018), its modifications modified LI-RADS (mLI-RADS) and revised LI-RADS (rLI-RADS), for diagnosing hepatocellular carcinoma (HCC) remains poorly understood and requires further validation. This multicentre study aimed to evaluate the diagnostic performance of three algorithms in diagnosing HCC.

We included 1092 untreated patients at risk for developing HCC who underwent gadoxetic acid-enhanced MRI across three independent cohorts from January 2020 to December 2022. Two readers independently interpreted each hepatic lesion and recorded their imaging features. The readers' judgements regarding whether the lesion was HCC or not were also noted. Non-HCC cases were confirmed based on histologic and clinical follow-up data, while HCC cases were pathologically confirmed. Diagnostic performance metrics were compared using bootstrap resampling and generalised estimating equations. Additionally, the diagnostic odds ratio (DOR) was evaluated.

Of 1313 lesions, 52.3% (687/1313) were diagnosed as HCC. For all hepatic lesions, mLI-RADS achieved higher sensitivity (82.9%) and accuracy (84.1%) than LI-RADS v2018 (sensitivity, 79.9%, p = 0.024; accuracy, 83.2%) and rLI-RADS (sensitivity, 81.3%; accuracy, 83.8%), while maintaining a similar positive predictive value (mLI-RADS, 86.2%; LI-RADS v2018, 86.9%; rLI-RADS, 86.9%). The DORs were 28.3 (95% CI: 21.1-38.0) for mLI-RADS, 27.8 (95% CI: 20.6-37.7) for LI-RADS v2018 and 26.0 (95% CI: 19.3-35.0) for rLI-RADS. The readers' judgement exhibited higher accuracy than that of three algorithms (87.7%: 83.2%-84.1%).

For the diagnosis of HCC in an HBV-predominant cohort, mLI-RADS showed higher performance compared with LI-RADS v2018 and rLI-RADS. Reader judgement achieved higher accuracy than all algorithms, highlighting the role of clinical expertise.

NCT06663904.