Improving rectal CT image quality with a deep learning image reconstruction algorithm(PDF)
《中国医学物理学杂志》[ISSN:1005-202X/CN:44-1351/R]
- Issue:
- 2024年第8期
- Page:
- 975-981
- Research Field:
- 医学影像物理
- Publishing date:
Info
- Title:
- Improving rectal CT image quality with a deep learning image reconstruction algorithm
- Author(s):
- QIAO Wenjun1; 2; ZHOU Fang1; LIU Quanfen1; HUANG Chantao1; XU Yikai1; 2
- 1. Department of Imaging Diagnostics, Nanfang Hospital, Southern Medical University, Guangzhou 510515, China 2. Medical Radiation Protection Committee of Radiation Protection Association of Guangdong Province, Guangzhou 510515, China
- Keywords:
- Keywords: rectum computed tomography deep learning image reconstruction image quality
- PACS:
- R318；R811.1
- DOI:
- DOI:10.3969/j.issn.1005-202X.2024.08.008
- Abstract:
- Abstract: Objective To improve the CT image quality of the anorectal junction in venous phase using a new deep learning image reconstruction (DLIR) algorithm. Methods A retrospective analysis was conducted on 71 patients undergoing pelvic computed tomography (CT) scans. All CT images were reconstructed at a thin slice thickness of 0.625 mm using 50% ASiR-V, low-, medium- and high-intensity DLIR (DLIR-L, DLIR-M and DLIR-H). The CT attenuations and standard deviation values of anal canal and hip fat were measured for each reconstruction group. With the standard deviation of hip fat as background noise, the contrast-to-noise ratio (CNR) and signal-to-noise ratio (SNR) of anal canal were calculated. Two radiologists independently assessed image quality and diagnostic confidence for local invasion of rectal cancer using the 5-point Likert scale. The objective measurement indicators and subjective scores were analyzed and compared, and Kappa test was used to evaluate the consistency. Results The differences in CT value of anal canal and hip fat among the groups were trivial (P>0.05), but fat SD, anal canal SNR and CNR (P<0.05) differed significantly, with lowest fat SD, highest anal canal SNR and CNR in DLIR-H group, while highest fat SD, lowest anal canal SNR and CNR in 50% ASiR-V group. Compared with 50% ASiR-V group, DLIR-H group decreased fat SD by 44.3%, but increased anal canal SNR and CNR by 89.5% and 92.1%, respectively (P<0.05). The subjective score of 4 groups were significantly different (P<0.05), decreasing from DLIR-H to 50% ASiR-V, and the inter-group differences were significant (P<0.05), except the difference between 50% ASiR-V group and DLIR-L group (P>0.05). There was a statistically significant difference in the diagnostic confidence for local invasion of rectal cancer among different groups (P<0.05), and the scores were significantly higher in DLIR-M and DLIR-H groups than in 50% ASiR-V and DLIR-L groups (P<0.05). Conclusion Compared with the standard 50% ASiR-V image, DLIR-M and DLIR-H reconstruction algorithms can effectively improve the image quality for the anorectal junction in CT imaging. The higher-intensity DLIR results in better image quality and stronger ability to display fine structures, which can provide more evidences for clinical precision evaluation and personalized precision treatment.
Last Update: 2024-08-31