深度学习重建算法在头颈部CTA成像中的应用价值
The Application Value of Deep Learning Image Reconstruction Algorithm in Head and Neck CT Angiography
林优优 1张秋爽 2潘璟琍 2丁建荣3
作者信息
- 1. 浙江省台州医院(浙江台州 317000);台州恩泽医疗中心(集团)恩泽医院(浙江台州 318050)
- 2. 浙江省台州医院(浙江台州 317000)
- 3. 浙江省台州医院(浙江台州 317000);台州市循证影像医学重点实验室(浙江临海 317000)
- 折叠
摘要
目的:探讨深度学习重建算法(DLIR)对头颈部CTA图像质量的影响.方法:回顾性收集50例头颈部CTA图像,分别进行DLIR-H、DLIR-M、DLIR-L和ASiR-V50％重建.测量并计算四组图像血管的背景噪声(SD)、血管锐利度(ERS)以及各重要层面血管的信噪比(SNR)和对比噪声比(CNR),并对四组图像进行主观质量评分.结果:随着DLIR强度增加,血管SD值显著降低(P＜0.05),四组图像的ERS无统计学差异(P＞0.05).除大脑中动脉M1段外,其他血管SNR均有统计学差异(P＜0.05),表现为DLIR-H最高,DLIR-M次之.所有血管CNR均有统计学差异(P＜0.05),由高到低依次为DLIR-H,DLIR-M,DLIR-L和ASiR-V50％.四组图像主观质量评分无统计学差异(P＞0.05).结论:与ASIR-V50％相比,DLIR-H可以显著提高头颈部CTA图像质量,为优化头颈部CTA扫描方案创造条件.
Abstract
Objective:To evaluate the impact of deep learning image reconstruction algorithm on the image quality in head and neck CT angiography.Methods:A total of 50 patients were retrospectively collected,all of the images were reconstructed by DLIR-H,DLIR-M,DLIR-L and ASiR-V50％.The background noise(standard deviation,SD),the edge rise slope(ERS),the signal-to-noise ratio(SNR),the contrast-to-noise ratio(CNR)were calculated and compared.Subjective evaluation of the four sets of images was also assessed by two radiologists.Results:As the strength of DLIR increased,the SD of vessels at each measurement level significantly decreased(P＜0.05).There was no statistically significant difference in ERS among the four groups of images(P＞0.05).The SNR of vessels showed statistical differences(P＜0.05),with DLIR-H being the highest,followed by DLIR-M(except for MCA-M1).The CNR of all vessels showed statistical differences(P＜0.05),with DLIR-H,DLIR-M,DLIR-L,and ASiR-V 50％in descending order.There was no statistically significant difference in subjective quality scores among the four groups of images(P＞0.05).Conclusion:Compared with ASiR-V 50％,DLIR-H can significantly improve the image quality in head and neck CT angiography and optimize head and neck CT angiography protocols.
关键词
CT血管成像/深度学习重建算法/图像质量Key words
CT angiography/deep learning image reconstruction algorithm/image quality引用本文复制引用
出版年
2024
NETL
NSTL
万方数据