摘要
目的探讨深度学习重建(deep-learning reconstruction,DLR)结合小视野(small field-of-view,sFOV)高分辨率扫描在提高手指磁共振成像(magnetic resonance imaging,MRI)图像质量中的价值。材料与方法前瞻性纳入33例健康志愿者和24例手部疾病患者,每位受检者均接受小视野高分辨率T2加权自旋回波成像矢状位序列(the small field-of-view high-resolution T2-weighted turbo spin-echo sequence,TSE-sFOV)和DLR结合TSE-sFOV(DLR combined with TSE-sFOV,TSE_(DL)-sFOV)的MRI扫描。采用4分法对57例样本的两组图像的整体图像质量(基于图像对比度、边缘锐利度、噪声、伪影)和解剖结构清晰度(包括骨、关节软骨、肌腱和韧带)进行主观评分;对其中24例样本的两组图像的病灶显示(包括病灶对比度及边缘锐利度、病灶位置及内部形态)和诊断置信度进行评分。评估57例样本两组图像的疾病检出能力(包括骨改变、关节间隙改变、肌腱异常、软组织异常)进行0或1检出。比较两组图像的信噪比(signal-to-noise ratio,SNR)、对比噪声比(contrast-to-noise ratio,CNR)。结果在主观评分中,57例样本的TSE_(DL)-sFOV组图像在整体图像质量、骨、关节软骨的评分中均高于TSE-sFOV组(P均<0.05),肌腱韧带方面的评分差异无统计学意义(P>0.05);在24例样本的病灶显示和诊断置信度方面,TSE_(DL)-sFOV组图像评分均高于TSE-sFOV组,差异具有统计学意义(P均<0.05)。在疾病检出能力方面,两组图像的疾病检出结果差异无统计学意义(P均>0.05);两组图像的一致性很好(kappa值均>0.84)。在客观评价中,TSE_(DL)-sFOV组的SNR和CNR均高于TSE-sFOV组(P均<0.05)。结论DLR结合sFOV手指MRI在缩短扫描时间的前提下,降低了噪声并提高了图像质量,为临床提供了更精准的影像依据。
Objective:To explore the value of small field-of-view(sFOV)high-resolution scanning based on deep-learning reconstruction(DLR)algorithm in improving the imaging quality of finger magnetic resonance imaging(MRI).Materials and Methods:Thirty-three healthy volunteers and 24 patients with hand diseases were prospectively recruited.Both the small field-of-view high-resolution T2-weighted spin-echo sequence(TSE-sFOV)and DLR combined with TSE-sFOV(TSE_(DL)-sFOV),were conducted on the subjects.A 4-point scale was used to subjectively evaluate the overall image quality(based on image contrast,edge sharpness,noise and artifact)and the clarity of anatomical structures(including bone,articular cartilage,tendon and ligament)in the two sets of images from 57 samples;Additionally,The lesion display(including lesion contrast and edge sharpness,lesion location and internal morphology)and diagnostic confidence were scored for 24 samples.The disease detection capabilities(including bone changes,joint space changes,tendon abnormalities,and soft tissue abnormalities)of the two groups of images from 57 samples were assessed as either 0 or 1.The signal-to-noise ratio(SNR)and contrast-to-noise ratio(CNR)of the two sets of images were compared.Results:In the subjective evaluation,the TSE_(DL)-sFOV group of images scored higher than the TSE-sFOV in overall image quality,bone and articular cartilage(P<0.05),while there was no statistical difference in tendon and ligament scores.For lesion display and diagnostic confidence in the 24 samples,the TSE_(DL)-sFOV group of images scored higher than the TSE-sFOV group,with statistical difference(P<0.05).In terms of disease detection capabilities,there was no statistical difference between the two groups of images(P>0.05),and the consistency between the two sets of images was excellent(Kappa>0.84).In the objective evaluation,the SNR and CNR of the TSE_(DL)-sFOV group of images were higher than those of the TSE-sFOV group(P<0.05).Conclusions:DLR combined with sFOV finger MRI can reduce the noise and improve the image quality under the premise of shortening scanning time.This provides more precise images for clinic.
作者
陆阿琴
徐露露
徐磊
郝绍伟
邹月芬
LU Aqin;XU Lulu;XU Lei;HAO Shaowei;ZOU Yuefen(Department of Radiology,the First Affiliated Hospital of Nanjing Medical University,Nanjing 210029,China;Siemens Healthineers Digital Technology(Shanghai)Co.,Ltd.,Shanghai 200000,China)
出处
《磁共振成像》
北大核心
2025年第7期52-57,共6页
Chinese Journal of Magnetic Resonance Imaging
关键词
手指
磁共振成像
深度学习
高分辨率
图像质量
finger
magnetic resonance imaging
deep learning
high resolution
image quality
