Deep learning-based and hybrid-type iterative reconstructions for CT: comparison of capability for quantitative and qualitative image quality improvements and small vessel evaluation at dynamic CE-abdominal CT with ultra-high and standard resolutions

Ryo Matsukiyo, Yoshiharu Ohno, Takahiro Matsuyama, Hiroyuki Nagata, Hirona Kimata, Yuya Ito, Yukihiro Ogawa, Kazuhiro Murayama, Ryoichi Kato, Hiroshi Toyama

Research output: Contribution to journalArticlepeer-review

1 Citation (Scopus)

Abstract

Purpose: To determine the image quality improvement including vascular structures using deep learning reconstruction (DLR) for ultra-high-resolution CT (UHR-CT) and area-detector CT (ADCT) compared to a commercially available hybrid-iterative reconstruction (IR) method. Materials and method: Thirty-two patients suspected of renal cell carcinoma underwent dynamic contrast-enhanced (CE) CT using UHR-CT or ADCT systems. CT value and contrast-to-noise ratio (CNR) on each CT dataset were assessed with region of interest (ROI) measurements. For qualitative assessment of improvement for vascular structure visualization, each artery was assessed using a 5-point scale. To determine the utility of DLR, CT values and CNRs were compared among all UHR-CT data by means of ANOVA followed by Bonferroni post hoc test, and same values on ADCT data were also compared between hybrid IR and DLR methods by paired t test. Results: For all arteries except the aorta, the CT value and CNR of the DLR method were significantly higher compared to those of the hybrid-type IR method in both CT systems reconstructed as 512 or 1024 matrixes (p < 0.05). Conclusion: DLR has a higher potential to improve the image quality resulting in a more accurate evaluation for vascular structures than hybrid IR for both UHR-CT and ADCT.

Original languageEnglish
Pages (from-to)186-197
Number of pages12
JournalJapanese journal of radiology
Volume39
Issue number2
DOIs
Publication statusPublished - 02-2021

All Science Journal Classification (ASJC) codes

  • Radiology Nuclear Medicine and imaging

Fingerprint Dive into the research topics of 'Deep learning-based and hybrid-type iterative reconstructions for CT: comparison of capability for quantitative and qualitative image quality improvements and small vessel evaluation at dynamic CE-abdominal CT with ultra-high and standard resolutions'. Together they form a unique fingerprint.

Cite this