基于深度学习算法的腰椎MRI临床应用研究  

Clinical Application of Lumbar Spine MRI Based on Deep Learning Algorithm

作　　者：胡英慧 赵文静[2] 王文娟[2] 王玺 隋心彤 王琦[2] HU Ying-hui;ZHAO Wen-jing;WANG Wen-juan;WANG Xi;SUI Xin-tong;WANG Qi(School of Medical Imaging,Shandong Second Medical University,Weifang 261053,Shandong Province,China;Department of Radiology,Weifang People's Hospital,Weifang 261041,Shandong Province,China)

机构地区：[1]山东第二医科大学医学影像学院,山东潍坊261053 [2]潍坊市人民医院放射科,山东潍坊261041

出　　处：《中国CT和MRI杂志》2024年第12期160-163,共4页Chinese Journal of CT and MRI

摘　　要：目的比较采用深度学习重建算法的腰椎快速自旋回波(FSE)序列图像和原始图像的图像质量。方法回顾性分析130名具有腰痛症状的患者,采用3.0T MRI进行腰椎2D FSE序列检查,包括矢状位T1WI、T2WI、T2WI-FS序列和横断位T2WI序列。一次扫描完成后,DLR算法引擎根据加速协议,生成原始图像(FSE0)和使用DLR处理后的图像(FSEDL)。两名放射科诊断医师对所有序列图像的整体图像质量、清晰度、解剖结构显示进行主观评分,并进行一致性检验。客观定量图像质量分析通过分别计算腰椎体和椎间盘的信噪比和对比噪声比来评价。结果总扫描时间为3分41秒。所有序列的FSEDL图像的腰椎椎体、椎间盘的SNR、CNR均高于FSE0图像(P均<0.05)。且FSEDL图像具有较高的整体图像质量和清晰度,解剖结构显示更加清晰(P均<0.05);两名评分者间的一致性为0.754-0.923之间。结论在腰椎常规2D FSE序列成像中,使用深度学习重建技术,能在4分钟内完成扫描的同时,提高总体图像质量。Objective To compare the image quality of lumber fast spin echo sequence images and original images by deep learning reconstruction algorithm.Methods130 patients with low back pain were analyzed retrospectively.Lumbar two dimensional(2D)fast spin echo(FSE)sequences was performed with 3.0T MRI,including T1-Weighted Image(T1WI),T2-Weighted Image(T2WI),T2-Weighted Fat Suppressed Image(T2WI-FS)sequence and transverse T2WI sequence.Once a scan is completed,the DL reconstruction algorithm engine will generate the original image(FSE0)and the image processed with DLR(FSEDL)according to the acceleration protocol.The overall image quality,clarity and anatomical structure of all sequence images were subjectively scored by two radiologists.And the consistency of the scores between the two physicians was tested.Objective quantitative image quality analysis was evaluated by calculating SNR and CNR of lumbar vertebrae and intervertebral discs respectively.Results The total scanning time was 3 minutes and 41 seconds.The SNR and CNR of lumbar vertebrae and intervertebral disc in all FSEDL images were higher than those in FSE0 images.And FSEDL images had higher overall image quality and sharpness,and the anatomical structure was more clearly displayed(P<0.05).The excellent consistency between the two raters was between 0.754 and 0.923.Conclusion In conventional lumbar 2D FSE sequence imaging,using the deep learning reconstruction technique can improve the overall image quality while scanning within 4 minutes.

关 键 词：腰椎 深度学习重建法 磁共振成像 信噪比 对比噪声比 

分 类 号：R445.2[医药卫生—影像医学与核医学]