Clinical Utility of Deep Learning-based Multiple Arterial Phase MRI in Hepatocellular Carcinoma.

Abstract

Purpose To evaluate the clinical utility of deep learning-based ultrafast multiple arterial phase (AP) MRI for diagnosing hepatocellular carcinoma (HCC), compared with conventional single AP imaging, using both extracellular agents (ECAs) and hepatobiliary agents (HBAs). Materials and Methods This prospective study included participants with suspected HCC who underwent either ECA- or HBA-enhanced MRI between September 2024 and March 2025. Outcomes included late AP capture rate, image quality (overall quality, motion artifacts, noise, liver and lesion edge sharpness, and lesion conspicuity), diagnostic performance (lesion, arterial phase hyperenhancement [APHE], and HCC detection rates), and hepatic arterial visualization scoring. Wilcoxon rank sum, Pearson χ², and Fisher exact tests compared characteristics from multiple and single AP MRI. Results The final analysis included 128 participants who underwent ECA-enhanced MRI (64 multiphase, 64 single-phase; median age, 61 years [IQR, 54-69 years]; 103 male) and 108 participants who underwent HBA-enhanced MRI (54 multiphase, 54 single-phase; median age, 62 years [IQR, 56-69 years]; 83 male). In the ECA group, multiple AP MRI resulted in greater late AP capture (98% vs 81%; P = .001); higher scores for overall image quality (P = .03), motion artifacts (P < .001), lesion edge sharpness (P < .001), and lesion conspicuity (P = .007); and higher lesion (98% vs 90%; P = .01), APHE (96% vs 88%; P = .03), and HCC (96% vs 81%; P < .001) detection rates. In the HBA group, multiple AP MRI examinations also resulted in greater late AP capture (98% vs 85%; P = .04); higher scores for overall image quality (P = .01), motion artifacts (P = .04), and lesion edge sharpness (P = .005); and higher lesion (97% vs 88%; P = .04) and APHE (97% vs 86%; P = .02) detection rates. Multiphase imaging consistently achieved satisfactory hepatic arterial visualization in both contrast agent groups (mean scores > 3 on a four-point Likert scale for all categories). Conclusion Deep learning-based ultrafast multiphase arterial MRI improved late AP capture, image quality, and HCC diagnosis and enabled reliable hepatic arterial visualization within a single scan compatible with ECA and HBA. Keywords: MR-Imaging, Abdomen/GI, Liver Supplemental material is available for this article. © RSNA, 2026.