Answers are grounded in this paper and cite their sources.
Introduction: Accurate non-invasive characterisation of focal liver lesions (FLLs) determines whether patients proceed to surveillance, biopsy, or treatment. Contrast-enhanced ultrasound (CEUS) offers real-time, radiation-free, purely intravascular imaging, whereas magnetic resonance imaging (MRI) provides multiparametric soft-tissue contrast but is costlier, slower, and less available. Head-to-head evidence comparing the two modalities, particularly within the Liver Imaging Reporting and Data System (LI-RADS) framework, remains fragmented. We aimed to compare the diagnostic accuracy of CEUS and MRI for the characterisation of FLLs. Methods: This review followed PRISMA-DTA and was registered in PROSPERO (CRD420261514484). PubMed, Europe PMC, and ClinicalTrials.gov were searched to 13 November 2026, supplemented by citation searching. Diagnostic accuracy studies (including sub-analyses of randomised trials and primary observational studies) of adults with FLLs that evaluated CEUS against a reference standard, preferably with paired MRI, were eligible. Risk of bias was assessed with QUADAS-2 and, for comparative studies, QUADAS-C. Sensitivity and specificity were pooled with a bivariate binomial–normal generalised linear mixed model; summary receiver operating characteristic (SROC) curves were derived from the hierarchical model, and CEUS was compared with MRI-based imaging using bivariate meta-regression on paired data. Positive and negative predictive values (PPV, NPV), likelihood ratios (LR+, LR−), diagnostic odds ratios (DOR), area under the SROC curve (AUC), and accuracy of lesion characterisation were estimated. Certainty was graded with GRADE for test accuracy. Results: Nine studies (2,408 lesions) were included; seven provided paired CEUS and MRI-based (contrast-enhanced MRI or combined CT/MRI LI-RADS) data. CEUS yielded a summary sensitivity of 0.823 (0.754–0.876), specificity of 0.931 (0.881–0.961), LR+ of 11.97 (6.97–20.57), LR− of 0.190 (0.136–0.265), DOR of 63.0 (34.5–114.9), and an AUC of 0.944. Pooled PPV and NPV of CEUS were 0.948 (0.906–0.972) and 0.747 (0.597–0.856), and overall accuracy was 0.854 (0.804–0.892). MRI-based imaging achieved a sensitivity of 0.768 (0.653–0.854), specificity of 0.919 (0.811–0.968), and AUC of 0.906. In paired studies, CEUS and MRI-based imaging did not differ in sensitivity (relative sensitivity 1.03 (0.90–1.16)) or specificity (relative specificity 1.01 (0.98–1.05)); bivariate meta-regression showed no modality effect (likelihood-ratio test P = 0.91). For non-HCC malignancy, CEUS LR-M was more sensitive but less specific than CT/MRI LR-M (pooled PPV 0.34 vs 0.84). Leave-one-out estimates for CEUS remained within 0.80–0.84 for sensitivity and 0.92–0.94 for specificity; Deeks' test showed no small-study effect for CEUS (P = 0.66). Certainty was moderate for CEUS specificity, low for CEUS sensitivity, and low to very low for comparative estimates. Discussion: CEUS combines high specificity with moderate sensitivity, so a positive (LR-5 or malignant) CEUS result substantially raises the probability of the target condition, whereas a negative result does not reliably exclude it. Its accuracy is statistically indistinguishable from MRI-based LI-RADS in the same lesions, although heterogeneity, threshold effects in MRI studies, verification and incorporation bias, and the frequent pooling of CT with MRI limit inference. CEUS LR-M over-captures hepatocellular carcinoma (HCC), while CT/MRI LR-M is more specific for non-HCC malignancy. Conclusion: CEUS is a valid, safe, and accessible alternative and complement to MRI for characterising FLLs, with comparable sensitivity and specificity, and is particularly useful as a problem-solving tool for indeterminate observations. Prospective, adequately powered, fully paired studies using MRI alone as comparator and uniform reference standards are required.
No takes yet. Share an insight, caveat, or question.
Idrus et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: