Improved delineation of the cystic artery using super-resolution deep learning reconstruction in contrast-enhanced abdominal computed tomography
Open Access
- 18 May 2026
- journal article
- Published by Springer Nature in Radiological Physics and Technology
- Vol. 19 (3) , 950-960
Abstract
This study aimed to evaluate the image quality and delineation of the cystic artery and related abdominal vessels using Super-Resolution Deep Learning Reconstruction (SR-DLR) on contrast-enhanced computed tomography (CT), compared with standard Deep Learning Reconstruction (DLR). This retrospective study included 60 patients who underwent contrast-enhanced abdominal CT with an arterial phase between April and June 2025. Images were reconstructed using both SR-DLR and DLR algorithms. Quantitative evaluation included CT attenuation values, image noise, contrast-to-noise ratio (CNR), full width at half maximum (FWHM), and edge rise distance and edge rise slope (ERD/ERS) for the superior mesenteric artery (SMA) and proper hepatic artery (PHA). Three radiologists independently performed qualitative assessments of cystic artery, PHA, SMA, and cystic duct delineation, as well as image noise, sharpness, and artifacts. SR-DLR showed significantly lower image noise and higher CT attenuation values and CNR than DLR ( p < 0.001). SR-DLR yielded higher ERS and lower ERD for the PHA and SMA compared to DLR ( p ≤ 0.006). Qualitative analysis demonstrated that all three readers rated cystic artery delineation as significantly better with SR-DLR than with DLR ( p ≤ 0.018). SR-DLR also achieved higher scores for image sharpness and diagnostic acceptability ( p ≤ 0.001). SR-DLR significantly improves the delineation and image quality of the cystic artery and adjacent vascular anatomy compared with DLR. While no significant improvement in cystic artery detection was observed, the enhanced image quality may better support clinical confidence in preoperative planning for cholecystectomy and interventional radiology procedures.Keywords
Funding Information
- University of Tokyo
This publication has 32 references indexed in Scilit:
- Impact of Super-Resolution Deep Learning Reconstruction on Low-Dose CT in Patients with Central Venous Catheter or Central Venous PortJournal of Imaging Informatics in Medicine, 2025
- Deep learning reconstruction enhances image quality in contrast-enhanced CT venography for deep vein thrombosisEmergency Radiology, 2025
- Efficacy of Fine-Tuned Large Language Model in CT Protocol Assignment as Clinical Decision-Supporting SystemJournal of Imaging Informatics in Medicine, 2025
- Improving detection performance of hepatocellular carcinoma and interobserver agreement for liver imaging reporting and data system on CT using deep learning reconstructionAbdominal Radiology, 2023
- The effect of deep learning reconstruction on abdominal CT densitometry and image quality: a systematic review and meta-analysisEuropean Radiology, 2021
- Deep learning reconstruction of drip-infusion cholangiography acquired with ultra-high-resolution computed tomographyAbdominal Radiology, 2020
- Deep learning reconstruction improves image quality of abdominal ultra-high-resolution CTEuropean Radiology, 2019
- The clinical anatomy of cystic artery variations: a review of over 9800 casesSurgical and Radiologic Anatomy, 2015
- Assessment and classification of cystic arteries with 64-detector row computed tomography before laparoscopic cholecystectomySurgical and Radiologic Anatomy, 2015
- Devastating and Fatal Complications Associated With Combined Vascular and Bile Duct Injuries During CholecystectomyArchives of Surgery, 2002