Deep learning-reconstructed hepatobiliary MRI: Enhancing biliary delineation, lesion margin sharpness, and small lesion detection.

OBJECTIVE

To evaluate the comparative performance of a deep learning-reconstructed T1-weighted volumetric interpolated breath-hold examination (DL VIBE) on hepatobiliary phase MRI against the conventional T1 VIBE in biliary tree visualization, lesion detection, and artifact suppression.

MATERIALS AND

METHODS

This retrospective study included 94 patients with suspected hepatic lesions (145 lesions, March-May 2025). All underwent gadoxetic acid-enhanced MRI at hepatobiliary phase using DL-VIBE (1.5 mm slice thickness) and conventional VIBE (3.0 mm). Blinded evaluations assessed biliary anatomy, lesion margins, and image quality through quantitative metrics (edge rise distance [ERD], CNR) and qualitative scoring.

RESULTS

DL-VIBE demonstrated superior biliary visualization with 163 % higher tertiary duct detection (2.45 vs. 0.93 branches, p < 0.001) and improved continuity scores (2.97 vs. 2.60, p < 0.001). Lesion margins showed significantly improved sharpness (ERD: 0.93 ± 0.50 mm vs. 1.51 ± 0.10 mm; P < 0.001), while lesion detection rates were significantly higher overall (97.2 % vs. 91.0 %; P = 0.005) and particularly for micro lesions (96.3 % vs. 87.6 %; P = 0.004). DL-VIBE reduced image noise (3.82 vs. 2.25, p < 0.001) and suppressed ringing artifacts (3.55 vs. 2.30, p < 0.001) while maintaining bile duct CNR gains (230.50 vs. 135.14, p < 0.001), though with marginally lower texture naturalness (2.85 vs. 3.45, p = 0.001).

CONCLUSION

DL-VIBE improves biliary microanatomy delineation, lesion margin resolution, and small lesion detection in gadoxetic acid-enhanced hepatobiliary MRI, suggesting clinical utility for precision imaging within standard breath-hold durations.

CLINICAL RELEVANCE STATEMENT: DL-reconstructed 1.5-mm HBP VIBE serves as an improved alternative to conventional 3-mm VIBE, offering enhanced tertiary bile duct visualization and subcentimeter lesion detection for better clinical decision-making.