JHEP Rep. 2026 Mar 4:101813. doi: 10.1016/j.jhepr.2026.101813. Online ahead of print.
ABSTRACT
BACKGROUND AND AIMS: There is a growing interest in reducing contrast medium use and the lengthy scan duration in liver imaging. This proof-of-concept study aimed to evaluate the feasibility of deep learning-based generation of synthetic three-dimensional liver contrast-enhanced multiphasic MRI exams that are similar to ground-truth exams in hepatocellular carcinoma and cirrhosis.
METHODS: MRI exams from hepatocellular carcinoma or cirrhosis patients at a single academic center were retrospectively collected. A three-dimensional cycle-consistent generative adversarial network was trained to generate synthetic three-dimensional T1-weighted contrast-enhanced multiphasic liver MRI exams, including arterial, portal venous, delayed, and hepatobiliary phases, using two precontrast T1-weighted and T2-weighted input phases. Quantitative performance evaluated similarity, error, and overlap metrics between synthetic and ground-truth exams. For the qualitative multi-reader study, three blinded radiologists assessed the ground-truth and synthetic MRI exams using a comprehensive questionnaire. Questionnaire tasks 1-5: visual Turing test (ground-truth vs synthetic nature), image quality, anatomic accuracy, disease diagnosability, artifacts, task 6: Liver Imaging Reporting and Data System features.
RESULTS: The study included 3,198 MRI phases from 533 MRI exams from 185 hepatocellular carcinoma (mean age, 62.1 years±9.7[SD]; 141 males) and 182 cirrhosis patients (54.4 years±10.0; 111 males). Synthetic MRI exams achieved high quantitative and qualitative similarity to ground-truth exams. Quantitative analysis demonstrated high structural similarity index (0.86±0.03), overlap (0.97±0.05), and low symmetric mean absolute percent error (0.63%±0.23%). Qualitative multi-reader study showed no significant difference in tasks 1-5 (p=0.06-0.50) and high performance metrics in task 6 (accuracy:0.76-0.86; precision:0.96-1.00) with moderate to perfect Fleiss's kappa interrater agreement (0.58-1.00, p<0.001).
CONCLUSIONS: Deep learning enabled generation of synthetic three-dimensional liver contrast-enhanced multiphasic MRI exams from precontrast sequences, achieving high quantitative and qualitative similarity to ground-truth images.
IMPACT AND IMPLICATIONS: This work demonstrates the early feasibility of generating high-quality, three-dimensional contrast-enhanced multiphasic liver MRI exams from precontrast sequences, with synthetic exams showing strong agreement with ground truth across quantitative metrics and key qualitative criteria, including the visual Turing test, image quality, disease diagnosability, anatomic accuracy, artifact severity, and hepatocellular carcinoma Liver Imaging Reporting and Data System features. Despite the model currently representing a proof of concept based on a moderate single-center dataset, with a need for larger multicenter studies and external validation, the results highlight the potential to transform liver MRI workflows by reducing contrast media costs and potential side effects, significantly shortening acquisition time-especially the prolonged 20-minute hepatobiliary phase-and improving accessibility for patients unable to tolerate contrast-enhanced MRI due to renal impairment, contrast agent allergy, or claustrophobia.
PMID:41887530 | DOI:10.1016/j.jhepr.2026.101813