Impact of deep learning image reconstruction algorithms on the quality of gastric cancer CT images
10.3969/j.issn.1002-1671.2025.11.029
- VernacularTitle:深度学习图像重建算法对胃癌CT图像质量的影响
- Author:
Ya WANG
1
;
Xia MA
;
Yun SHEN
;
Yanbing YANG
;
Dazhi CHEN
;
Jinhua WU
Author Information
1. 宁夏回族自治区人民医院(宁夏医科大学附属自治区人民医院)医学影像中心,宁夏 银川 750004
- Publication Type:Journal Article
- Keywords:
stomach neoplasms;
deep learning;
image reconstruction;
computed tomography;
image quality
- From:
Journal of Practical Radiology
2025;41(11):1891-1894
- CountryChina
- Language:Chinese
-
Abstract:
Objective To explore the value of deep learning image reconstruction(DLIR)by comparing subjective and objec-tive evaluation of DLIR and adaptive statistical iterative reconstruction(ASIR-V)images in gastric cancer CT.Methods Abdominal CT images in the venous phase of 80 untreated patients with primary gastric cancer were included,and five CT reconstruction meth-ods of 50％ASIR-V,80％ASIR-V,DLIR-low(DLIR-L),DLIR-medium(DLIR-M)and DLIR-high(DLIR-H)were adopted,respec-tively.The objective evaluation included background standard deviation(SD),signal-to-noise ratio(SNR),contrast-to-noise ratio(CNR)values and pericancerous fat density resolution.The subjective evaluation included SD,overall image quality,the display of gastric cancer lesions,and diagnostic confidence in whether the serosal surface of the gastric wall was infiltrated.Results The subjective and objective evalua-tion indicators showed statistically significant differences among the five reconstruction models(P＜0.001).In terms of objective evalua-tion,the SD value of gastric cancer lesions in the DLIR-H was the lowest,while the SNR and CNR values were the highest among the five groups.In terms of subjective evaluation,the DLIR-M had the highest scores in gastric cancer lesions display,overall image quality and diagnostic confidence among the five groups.Conclusion Compared with ASIR-V,DLIR can significantly reduce image noise and improve image quality,and DLIR-M and DLIR-H are respectively the optimal subjective and objective reconstruction models for showing gastric cancer lesions.
