深度学习重建算法在磁共振颅脑增强中减少钆喷酸葡胺使用剂量的有效性研究  

Investigating the feasibility of reducing the usage of Gd-DTPA in MRI brain enhancement by improving the quality of acquired image through DL-Recon

作　　者：梁丹 张默[1,2] 马素文 卢洁[1,2] LIANG Dan;ZHANG Mo;MA Suwen;LU Jie(Department of Radiology and Nucler Medicine,Xuanwu Hospital Capital Medical University,Beijing 100053,China;Beijing KeyLaboratory of Magnetic Resonance Imaging and Brain Informatics,Beijing 100053,China)

机构地区：[1]首都医科大学宣武医院放射与核医学科,北京100053 [2]磁共振成像脑信息学北京市重点实验室,北京100053

出　　处：《磁共振成像》2023年第11期136-141,共6页Chinese Journal of Magnetic Resonance Imaging

基　　金：汇智人才工程-支持计划-领军人才项目(编号:HZ2021ZCLJ005)。

摘　　要：目的探讨通过应用深度学习重建(deep learning reconstruction,DL-Recon)算法提高颅脑MRI图像质量,实现在MRI临床实践中减少对比剂——钆喷酸葡胺(Gd-diethylenetriamine pentametric acid,Gd-DTPA)使用剂量的可行性。材料与方法收集2022年8月1日至2022年9月30日于我院放射科进行头部MRI增强扫描的60例患者。采用随机对照法,按照是否降低Gd-DTPA注射量平均分为正常药量组和降低药量组,并对两组患者分别使用常规重建方法和基于DL-Recon技术获取相应的T1WI图像。由两名医师在双盲条件下圈定感兴趣区进行信噪比(signal-to-noise ratio,SNR)及对比噪声比(contrast-to-noise ratio,CNR)分析,并对图像质量、伪影、均匀度及增强效果进行主观和客观评价,主观评价应用LIKERT五分制法,客观评价中通过分别计算图像背景以及额上回(superior frontal gyrus,SFG)、蛛网膜下腔(subarachnoid space,SAS)和红核(red nucleus,RN)的SNR对图像对比度进行评价。结果Gd-DTPA正常药量组中,相较于采用常规重建方法获得的图像,应用DL-Recon方法获取的图像在图像质量上显著升高(SNRSFG升高48.9%,CNRSFG提高91.5%);DL-Recon获取的图像伪影和图像整体质量得分均显著高于常规重建的图像(P<0.05)。降低药量组中,通过DL-Recon获取的图像与正常药量组的常规重建图像增强效果差异无统计学意义(P>0.05)。结论基于DL-Recon算法的MRI具备在保证图像质量的前提下降低Gd-DTPA注射量的应用能力。Objective:To investigate the feasibility of reducing the effect of Gd-diethylenetriamine pentametric acid(Gd-DTPA)dose on image quality in MRI brain enhancement by improving the acquired image through deep learning reconstruction(DL-Recon)algorithm.Materials and Methods:The patients were divided equally into two groups,the normal dose group and the reduced dose group,and the corresponding T1WI were acquired using conventional reconstruction methods and DL-Recon techniques for the two groups.The region of interest was determined for signal-to-noise ratio(SNR)and contrast-to-noise ratio(CNR)analysis under double-blind conditions by two associate chief,and subjective and objective evaluation of image quality,artefacts,homogeneity and enhancement effects were performed,with the subjective evaluation method based on the LIKERT guideline.The objective evaluation was performed by calculating the SNR of the image background,the superior frontal gyrus(SFG),the subarachnoid space(SAS)and the red nucleus(RN),respectively.The SNR was evaluated for image contrast.Results:The image quality obtained using DL-Recon was significantly increased(SNRSFG increased by 48.9%,CNRSFG increased by 91.5%)compared with that of conventional reconstruction method after injection of normal dose,and there was no significant correlation with the injection dose of Gd-DTPA.The artifacts and overall quality scores of DL-Recon were significantly higher than those of conventional images(P<0.05).There was no significant difference in the enhancement effect of DL-Recon image+reduced dose group compared with that of conventional reconstruction in the normal injection dose group(P>0.05).Conclusions:MRI DL-Recon algorithm has the ability to reduce the injection amount of Gd-DTPA on the premise of ensuring image quality.

关 键 词：颅内肿瘤 钆喷酸葡胺 低剂量 深度学习重建算法 图像质量 磁共振成像 

分 类 号：R445.2[医药卫生—影像医学与核医学] R739.41[医药卫生—诊断学]