Back to all papers

Deep learning reconstruction combined with contrast enhancement boost improves image quality of low-contrast-dose ctpa: a prospective randomized study.

October 30, 2026pubmed logopapers

Authors

Xie D,Lin J,Huang M,Lai Z,Yang X,Xu R,Cai H,Ma H

Affiliations (5)

  • Department of Radiology, The First Affiliated Hospital, Sun Yat-sen University, No. 58, Zhongshan Road 2, Guangzhou, 510080, Guangdong, China.
  • Department of Radiology, The First Affiliated Hospital, Sun Yat-sen University, No. 58, Zhongshan Road 2, Guangzhou, 510080, Guangdong, China. [email protected].
  • Research Collaboration, Canon Medical Systems, No.10 Huaxia Road, Guangzhou, 510623, Guangdong, China.
  • Department of Radiology, The First Affiliated Hospital, Sun Yat-sen University, No. 58, Zhongshan Road 2, Guangzhou, 510080, Guangdong, China. [email protected].
  • Department of Radiology, The First Affiliated Hospital, Sun Yat-sen University, No. 58, Zhongshan Road 2, Guangzhou, 510080, Guangdong, China. [email protected].

Abstract

High-quality CT pulmonary angiography (CTPA) is vital for pulmonary embolism diagnosis. This study evaluated the effect of combining deep learning reconstruction (DLR) with contrast enhancement boost (CE-boost) in low-contrast-dose CTPA. In this prospective randomized trial (September-December 2023), 106 patients were assigned to a low-contrast-dose (3 mL/s, 30 mL; Group A) or standard-contrast-dose (4 mL/s, 50 mL; Group B) CTPA protocol. Four image datasets were generated from each case in Group A using a 2 × 2 factorial design: A1, hybrid iterative reconstruction (HIR); A2, DLR (AiCE); A3, HIR plus CE-boost; A4, DLR plus CE-boost. Group B used HIR. Quantitative (CT attenuation, noise, SNR, CNR) and qualitative (overall image quality, subsegmental artery visibility) endpoints were assessed with linear mixed models and ART ANOVA with Holm correction. Each group comprised 53 patients (mean age: 59 ± 15 vs. 58 ± 18 years; 27 vs. 26 women). DLR significantly reduced noise and improved SNR/CNR (all Holm-corrected p < 0.001). CE-boost significantly increased CT attenuation, SNR, and CNR (all p < 0.001) without affecting noise. A significant DLR × CE-boost interaction was observed for CNR in all ROIs (all p < 0.05), indicating synergy. Qualitatively, DLR improved both overall quality and visibility, whereas CE-boost improved visibility only. The combined protocol (A4) achieved the highest CNR and visibility among all four datasets, significantly surpassing both the baseline low-contrast-dose (A1) and standard-contrast-dose (Group B) protocols (all Bonferroni-corrected p < 0.0083/0.025). Combining DLR with CE-boost synergistically optimizes low-contrast-dose CTPA, enabling a 40% reduction in iodine load and a 25% reduction in injection rate while enhancing image quality beyond that of standard-contrast-dose imaging. This protocol may offer a viable strategy for patients requiring strict contrast limitation.

Topics

Deep LearningContrast MediaPulmonary EmbolismComputed Tomography AngiographyRadiographic Image Interpretation, Computer-AssistedJournal ArticleRandomized Controlled Trial

Ready to Sharpen Your Edge?

Subscribe to join 11k+ peers who rely on RadAISlice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.