基于压缩感知的全脑三维同时非增强血管成像和斑块内出血成像研究

An imaging study of three-dimensional simultaneous non-contrast angiography and intraplaque hemorrhage of whole brain based on compressed sensing

目的采用不同加速因子(accelerated factors,AF)优化头部基于压缩感知的三维同时非增强血管成像和斑块内出血成像(three-dimensional simultaneous non-contrast angiography and intraplaque hemorrhage based on compressed sensing,3D CS SNAP),以期在短时间内获得全脑覆盖的高质量血管成像(magnetic resonance angiography,MRA)和血管壁图像。材料与方法前瞻性招募30名健康志愿者,10名志愿者进行头部三维同时非增强血管成像和斑块内出血成像(three-dimensional simultaneous non-contrast angiography and intraplaque hemorrhage,3D SNAP)参数的预扫描及选择;20名志愿者在优化后的预扫描序列基础上,行头部的常规3D SNAP以及不同AF(2、2.5、3、4、5)的3D CS SNAP序列扫描。两名影像科医生在原始图像上划定血管腔、血管壁、邻近白质的感兴趣区,测量信号强度和标准差,并计算血管腔的信噪比(signal to noise ratio,SNR)、对比噪声比(contrast to noise ratio,CNR)、血管壁的对比组织比(contrast-to-tissue ratio,CTR)和CTR效率(CTR efficiency,CTReff);同时,对不同AF的3D CS SNAP图像质量进行主观评分。使用Kappa检验和组内相关系数(intraclass correlation coefficient,ICC)分析两名观察者的组间一致性;对比分析不同AF间测量数据以及主观评分的差异。结果两名观察者图像质量主客观评价一致性良好(主观评分Kappa:0.568~0.884,数据测量ICC:0.602~0.968)。当CS AF≥4时,血管腔SNR、CNR及主观评价较常规3D SNAP的差异有统计学意义(P<0.05);当CS AF=2.5、3、4、5时,血管壁CTReff与常规3D SNAP差异均有统计学意义(P<0.05)。结论3D CS SNAP技术可以在相对更短的时间里获得全脑MRA和血管壁图像,一次扫描同时提供血管形态和血管壁斑块的诊断信息。推荐AF为2应用于轻中症患者,AF为3应用于重症患者。

Objective:To optimize three-dimensional simultaneous non-contrast angiography and intraplaque hemorrhage(3D SNAP)based on compressed sensing(CS 3D SNAP),we performed different acceleration factors(AF)to obtain high-quality magnetic resonance angiography(MRA)and vascular wall images capable of whole brain coverage in a short time.Material and Methods:Thirty healthy volunteers were prospectively recruited in this study.Ten volunteers were performed the pre-scan for selecting a precise head three-dimensional simultaneous non-contrast angiography and intraplaque hemorrhage(3D SNAP)parameter.Twenty volunteers were scanned with conventional 3D SNAP and CS 3D SNAP(AF=2,2.5,3,4,or 5)on the basis of the optimized pre-scan sequence.Two radiologists delineated the region of interest of vascular lumen,vascular wall,and adjacent white matter on the original images,then measured the signal intensity and the standard deviation.The signal to noise ratio(SNR)and the contrast to noise ratio(CNR)of vascular lumen,the contrast-to-tissue ratio(CTR)and the CTR efficiency(CTReff)of the vascular wall were further calculated.The image quality of each dataset were scored subjectively by two radiologists.The consistency of measurements and subjective scores of the two observers were analyzed using Kappa and intra-class correlation coefficient(ICC)test.Compare and analyze the differences in measurement parameters and subjective scores between different AF.Results:The measured data and subjective scores of the two observers were in good consistent(Kappa:0.568-0.884,ICC:0.602-0.968).When CS AF is greater than or equal to 4,the SNR,CNR of vascular lumen and subjective scores were significantly different from those of conventional 3D SNAP(P<0.05).When CS AF is equal to 2.5,3,4,or 5,the CTReff of vascular wall were significantly different compared with conventional 3D SNAP(P<0.05).Conclusions:3D CS SNAP can obtain whole cerebrovascular MRA and vascular wall images in a relatively short period of time,and a single scan provides diagnostic information on both vascular morphology and plaque in the vessel wall.AF of 2 is recommended for mild and moderate disease patients and AF of 3 is recommended for critical patients.