张浩南; 宋清伟; 张楠; 宋宇; 浦仁旺; 王楠; 刘爱连

doi:10.12015/issn.1674-8034.2023.02.022

点赞 0
分享 0
收藏 0
纠错

• 技术研究 •

压缩感知技术在腰椎磁共振快速成像中的应用

磁共振成像, 2023,14(2) : 132-137,144. DOI: 10.12015/issn.1674-8034.2023.02.022

摘要

目的

探讨压缩感知（compressed sensing, CS）不同加速因子（acceleration factor, AF）对腰椎MRI图像质量的影响。

材料与方法

招募32例受检者（男12例，女20例），年龄（45.28±14.11）岁。行3.0 T MR结合敏感度编码（sensitivity encoding, SENSE）以及CS的腰椎矢状位T1WI、T2WI和轴位T2WI序列扫描，采用的AF分别为无加速、SENSE AF=2、CS AF=2、3、4、5。两观察者在矢状位T1WI、T2WI和轴位T2WI划定感兴趣区测量信号强度（signal intensity, SI）和噪声强度（standard deviation, SD），并计算信噪比（signal to noise ratio, SNR）和对比噪声比（contrast to noise ratio, CNR）。对图像质量进行五分法主观评分。组内相关系数（intra-class correlation coefficient, ICC）、Kappa检验分析两观察者测量数据和主观评分的一致性。对各序列不同组间图像的SNR、CNR和主观评分采用单因素ANOVA分析。

结果

两观察者测量数据和主观评分一致性良好（ICC：0.878～0.997，Kappa：0.763～0.948）。单因素ANOVA检验矢状位T1WI、T2WI、轴位T2WI不同AF间SNR、CNR及评分差异有统计学意义（P＜0.01）。两两比较结果：当CS=4时，矢状位T1WI椎体和间盘的SNR、矢状位T2WI椎体和间盘的SNR、CNR和主观评分较常规序列差异有统计学意义（P＜0.05）；当CS=3时，轴位T2WI椎体SNR较常规序列差异有统计学意义（P＜0.05）。

结论

腰椎MRI扫描时间随着AF的增加逐渐降低，在保证图像质量的前提下，临床推荐AF分别以3、3、2行腰椎矢状位T1WI、T2WI和轴位T2WI序列扫描。

引用本文: 张浩南, 宋清伟, 张楠, 等. 压缩感知技术在腰椎磁共振快速成像中的应用 [J] . 磁共振成像, 2023, 14(2) : 132-137,144. DOI: 10.12015/issn.1674-8034.2023.02.022.

参考文献导出: Endnote NoteExpress RefWorks NoteFirst 医学文献王

扫描看全文

正文

作者信息

基金 0 关键词 0

English Abstract

阅读 0 评论 0

相关资源

引用 | 论文 | 视频

本刊刊出的所有论文不代表本刊编委会的观点，除非特别声明

0 前言

MRI因良好的软组织对比度、较高空间分辨率、无创伤、无电离辐射等优势^{[1, 2]}，被认为是腰椎及间盘病变的首选影像学检测^{[3, 4, 5]}。在扫描检查期间，患者必须长时间保持静止以防止运动伪影^[6]。尤其对于有严重腰椎病变患者，难以配合完成检查。因此，在更短时间内获得满足临床诊断需求的图像是目前腰椎及间盘MRI检查所面临的重要问题。传统敏感度编码（sensitivity encoding, SENSE）利用多通道相控阵线圈并行采集减少相位编码的数目，从而缩短扫描时间。但SENSE采集过程中，会出现幻像伪影、混叠伪影、卷积伪影，影响图像质量^[7]。压缩感知（compressed sensing, CS）利用随机化稀疏采样，采样频率远小于奈奎斯特采样定律^{[8, 9]}，应用于腰椎MRI可以显著缩短扫描时间，减少图像伪影^{[10, 11]}。以往研究将CS应用于腰椎三维（three dimensional, 3D）T2WI序列以及二维（two dimensional, 2D）矢状位梯度回波（gradient recalled echo, GRE）和快速自旋回波（turbo spin echo, TSE）T2WI序列^{[12, 13]}可行性良好。然而，临床上2D矢状位T1WI、T2WI和轴位T2WI应用更加广泛。本研究首次将CS应用于腰椎2D序列中，旨在探讨不同加速因子（acceleration factor, AF）对2D腰椎MRI图像质量与扫描时间的影响，并优化AF。

1 材料与方法

1.1 一般资料

于2021年6月至2022年3月在大连医科大学附属第一医院前瞻性招募32例（男12例，女20例）受检者，行腰椎MRI检查，年龄（45.28±14.11）岁。入组标准：（1）无椎体或盆腔恶性肿瘤、转移瘤及放化疗病史；（2）半年内无腰背部手术史。排除标准：（1）怀疑新发腰椎压缩骨折者；（2）严重脊柱后凸，难以卧床者；（3）腰椎钛合金内固定患者；（4）有MRI检查禁忌证者，如体内有起搏器、支架、金属植入物及患有幽闭恐惧症等。本研究遵照《赫尔辛基宣言》，经大连医科大学附属第一医院伦理委员会批准（批件号：PJ-KS-KY-2021-10），所有被试同意并签署知情同意书。

1.2 检查方法

使用Philips Ingenia CX 3.0 T MRI（Philips Healthcare, Best, the Netherlands）设备，12通道内置体线圈。所有被试均行常规和5种不同AF的腰椎矢状位TSE-T1WI和T2WI、轴位TSE-T2WI序列，具体为SENSE AF=2以及CS AF=2、3、4、5（图1）。（1）矢状位T1WI扫描参数：TR 324 ms，TE 9 ms，体素0.89 mm×2.01 mm×4.00 mm，层间距1 mm，信号平均采集次数（number of signal average, NSA）1，无加速、SENSE 2、CS 2、CS 3、CS 4、CS 5扫描时间分别为162.0 s、84.0 s、84.0 s、58.3 s、45.4 s、37.6 s；（2）矢状位T2WI扫描参数：TR 1600 ms，TE 90 ms，体素0.80 mm×1.83 mm×4.00 mm，层间距1 mm，NSA 1，无加速、SENSE 2、CS 2、CS 3、CS 4、CS 5扫描时间分别为103.0 s、51.3 s、51.3 s、35.2 s、25.6 s、22.4 s；（3）轴位T2WI扫描参数：TR 3200 ms，TE 129 ms，体素0.60 mm×0.95 mm×4.00 mm，层间距0.4 mm，NSA 2，无加速、SENSE 2、CS 2、CS 3、CS 4、CS 5扫描时间分别为216.0 s、115.0 s、115.0 s、76.0 s、64.0 s、50.9 s。

点击查看大图

图1

男，24岁，腰椎第5间盘退化。图1A～1F分别为加速因子（AF）为无加速、敏感度编码（SENSE）AF=2和压缩感知（CS）AF=2、3、4、5的矢状位T1WI-快速自旋回波（TSE）序列；1G～1L分别为AF为无加速、SENSE AF=2和CS AF=2、3、4、5的矢状位T2WI-TSE序列；1M～1R分别为AF为无加速、SENSE AF=2和CS AF=2、3、4、5的轴位T2WI-TSE序列。可见随着AF增大，噪声逐渐增大。

Fig. 1

Male, 24 years old, lumbar disc degeneration. 1A-1F show sagittal T1WI turbo spin echo (TSE) sequences with acceleration factor (AF) being non accelerated, sensitivity encoding (SENSE) AF=2, compressed sensing (CS) AF=2, 3, 4, and 5 respectively; 1G-1L show sagittal T2WI-TSE sequences with AF being non accelerated, SENSE AF=2, CS AF=2, 3, 4, and 5, respectively; 1M-1R show respectively axial T2WI-TSE sequences with AF being no acceleration, SENSE AF=2, CS AF=2, 3, 4, and 5。It can be seen that with the increase of AF, the noise gradually increases.

点击查看大图

图1

Fig. 1

1.3 数据测量及分析

1.3.1 数据测量

将扫描所得图像传至ISP（Philips, IntelliSpace Portall Version7）工作站，由两名观察者（观察者1为从事MRI工作5年的主管技师，观察者2为从事MRI工作4年的技师）在矢状位T1WI、T2WI椎体最大层面划定感兴趣区（region of interest, ROI），测量L1～L5椎体、间盘信号强度（signal intensity, SI）和噪声强度标准差（standard deviation, SD），ROI大小分别为140～150 mm²、40～50 mm²，并在L3～L5水平划定三个ROI测量脑脊液SI和SD，大小20～30 mm²，取平均值。在轴位T2WI划定ROI测量L1～L5椎体、间盘SI和SD值，ROI大小140～150 mm²，并在L3～L5水平划定ROI测量右侧腰大肌SI和SD，大小40～50 mm²（图2），取平均值。计算信噪比（signal to noise ratio, SNR）和对比噪声比（contrast to noise ratio, CNR）^[5]。矢状位T1WI、T2WI：SNR=SI_{椎体或间盘}/SD_脑脊液，CNR=（SI_{椎体或间盘}-SI_脑脊液）/SD_脑脊液；轴位T2WI：SNR=SI_{椎体或间盘}/SD_腰大肌，CNR=（SI_{椎体或间盘}-SI_脑脊液）/SD_腰大肌。

点击查看大图

图2

男，24岁，L5椎间盘囊肿。2A：在矢状位T2WI-快速自旋回波（TSE）序列椎体最大层面划感兴趣区测量椎体、间盘和脑脊液的信号强度（SI）和噪声强度（SD）；2B：在轴位T2WI-TSE序列间盘划感兴趣区测量间盘和右侧腰大肌的SI和SD。

Fig. 2

Male, 24 years old, L5 intervertebral disc cyst. 2A: Region of interest (ROI) is drawn on the largest section of sagittal T2WI-TSE sequence vertebral body to measure the signal intensity (SI) and standard deviation (SD) of vertebral body, intervertebral disc and cerebrospinal fluid; 2B: Diagram measure SI and SD of the intervertebral disc and right psoas major muscle by drawing ROI in axial T2WI-TSE sequence.

点击查看大图

图2

Fig. 2

1.3.2 主观评分

由两名观察者根据椎体间盘终丝马尾形态及边界清晰度、图像运动伪影大小以及诊断价值用五分法对图像进行评分（表1）。4分及4分以上满足诊断要求。

点击查看表格

表1

主观评分标准表

Tab. 1

Table of subjective scoring standards

表1

主观评分标准表

Tab. 1

Table of subjective scoring standards

分数	椎体间盘终丝马尾可见度	伪影大小	诊断价值
1	无法区分	运动伪影显著	无法诊断
2	可见度较差	运动伪影明显	诊断价值有限
3	可见度一般	运动伪影较明显	诊断价值一般
4	可见度良好	运动伪影较小	诊断价值较高
5	细节可见度高	无运动伪影	诊断价值高

1.4 统计学分析

使用统计软件SPSS 21.0对数据进行分析。使用组内相关系数（intra-class correlation coefficient, ICC）、Kappa检验分析两观察者测量数据及主观评分的一致性，结果≥0.75为一致性良好。对各序列不同组间图像的SNR、CNR和主观评分采用单因素ANOVA分析，P＜0.05为差异具有统计学意义。

2 结果

2.1 主观评分一致性检验

两观察者测量数据和主观评分一致性良好（ICC：0.878～0.997，Kappa：0.763～0.948；表2）。选择观察者1的测量数据和主观评分进行后续分析。

点击查看表格

表2

两观察者测量数据及主观评分一致性比较

Tab. 2

Comparison of the consistency of measurement data and subjective scores between the two observers

表2

两观察者测量数据及主观评分一致性比较

Tab. 2

Comparison of the consistency of measurement data and subjective scores between the two observers

	常规			SENSE 2			CS 2
	观察者1	观察者2	ICC/Kappa值	观察者1	观察者2	ICC/Kappa值	观察者1	观察者2	ICC/Kappa值
矢状位T1WI
SNR椎体	45.45±8.44	45.39±9.44	0.950	43.00±9.92	42.70±9.86	0.978	46.26±10.05	46.18±10.75	0.924
CNR椎体	31.41±8.08	31.38±8.97	0.973	29.86±9.03	29.77±9.08	0.984	32.42±9.39	32.28±10.02	0.947
SNR间盘	25.89±3.73	26.04±3.85	0.903	24.32±3.78	24.29±3.61	0.972	26.51±4.63	26.71±5.02	0.932
CNR间盘	11.85±3.13	12.04±3.18	0.968	11.18±3.09	11.36±3.05	0.990	12.67±3.63	12.82±3.78	0.975
评分	4.88±0.34	4.84±0.37	0.871	4.88±0.34	4.84±0.37	0.871	4.84±0.37	4.81±0.40	0.890
矢状位T2WI
SNR椎体	17.99±5.59	17.91±5.49	0.980	16.63±5.55	16.25±5.19	0.987	18.16±6.39	18.61±5.86	0.961
CNR椎体	26.21±7.40	26.40±7.45	0.968	24.38±7.84	24.21±7.77	0.987	25.22±9.05	26.15±8.70	0.977
SNR间盘	10.98±3.71	11.10±3.85	0.977	10.10±3.51	10.13±3.64	0.988	11.11±4.42	11.66±4.33	0.946
CNR间盘	33.22±8.49	33.22±8.33	0.966	30.92±8.55	30.34±8.19	0.980	32.27±10.44	33.10±9.87	0.976
评分	4.88±0.34	4.91±0.30	0.840	4.81±0.40	4.88±0.34	0.765	4.81±0.40	4.81±0.40	0.795
横断位T2WI
SNR椎体	20.74±4.46	20.89±4.72	0.975	19.36±4.28	19.41±4.50	0.987	22.05±6.33	22.49±5.96	0.979
CNR椎体	16.37±4.63	16.85±4.80	0.817	15.63±4.80	16.02±4.42	0.938	17.93±4.63	18.83±5.72	0.927
SNR间盘	20.93±7.77	21.17±7.81	0.987	19.73±8.03	19.79±8.16	0.995	22.30±8.80	23.06±9.26	0.983
CNR间盘	16.56±8.26	17.22±7.99	0.946	16.16±8.26	16.41±8.10	0.990	18.19±8.10	19.40±9.28	0.964
评分	4.91±0.30	4.88±0.34	0.840	4.84±0.37	4.84±0.37	0.763	4.84±0.37	4.81±0.40	0.890
	CS 3			CS 4			CS 5
	观察者1	观察者2	ICC/Kappa值	观察者1	观察者2	ICC/Kappa值	观察者1	观察者2	ICC/Kappa值
矢状位T1WI
SNR椎体	42.30±9.02	42.14±9.28	0.980	40.72±8.75	40.51±9.69	0.972	33.66±6.38	34.32±7.46	0.970
CNR椎体	29.46±8.36	29.36±8.43	0.985	28.54±8.07	28.45±8.61	0.979	23.76±6.21	24.28±6.99	0.981
SNR间盘	24.34±3.26	24.38±3.50	0.966	23.35±3.48	23.50±4.33	0.959	19.33±2.95	19.76±2.90	0.950
CNR间盘	11.54±2.86	11.60±3.06	0.985	11.17±3.00	11.45±3.39	0.974	9.45±2.47	9.77±2.45	0.974
评分	4.78±0.42	4.78±0.42	0.817	4.75±0.44	4.75±0.44	0.833	3.28±0.63	3.16±0.68	0.787
矢状位T2WI
SNR椎体	16.48±4.96	16.34±4.98	0.981	14.57±4.79	14.60±5.12	0.982	13.17±4.77	13.27±5.22	0.965
CNR椎体	24.59±7.28	24.98±7.44	0.971	22.06±7.01	22.36±6.70	0.985	19.54±5.93	19.61±6.64	0.909
SNR间盘	10.17±3.49	10.24±3.66	0.985	9.04±3.31	9.20±3.40	0.986	8.12±2.72	8.10±3.16	0.878
CNR间盘	30.90±8.25	31.09±8.43	0.968	27.59±7.46	27.75±7.40	0.979	24.59±7.76	24.78±8.59	0.949
评分	4.78±0.42	4.72±0.46	0.834	4.34±0.70	4.41±0.67	0.788	3.00±0.72	3.06±0.67	0.795
横断位T2WI
SNR椎体	18.30±4.36	17.49±3.79	0.963	10.83±3.12	11.03±3.11	0.959	8.10±1.78	8.07±1.65	0.980
CNR椎体	14.45±5.06	14.23±3.89	0.896	7.96±2.86	8.10±2.90	0.965	5.26±1.62	5.33±1.60	0.963
SNR间盘	18.77±8.29	17.95±7.38	0.987	11.10±5.54	11.21±5.22	0.983	8.34±4.52	8.34±4.46	0.997
CNR间盘	15.21±8.25	14.69±7.36	0.985	8.40±5.21	8.29±5.18	0.981	5.67±4.36	5.60±4.39	0.989
评分	4.75±0.44	4.75±0.44	0.833	4.34±0.70	4.41±0.67	0.788	2.03±0.70	2.06±0.67	0.948

注：SENSE为敏感度编码；CS为压缩感知；2、3、4、5分别为加速因子值；SNR为信噪比；CNR为对比噪声比。

2.2 两组ADC值及主观评分的差异性检验

ANOVA检验矢状位TSE-T1WI、T2WI、轴位TSE-T2WI不同AF间椎体和间盘的SNR、CNR及主观评分差异有统计学意义（P＜0.05；表3）。两两比较结果：当CS=4时，矢状位T1WI椎体和间盘的SNR、矢状位T2WI椎体和间盘的SNR、CNR和主观评分较常规序列差异有统计学意义（P＜0.05）；当CS=3时，轴位T2WI椎体SNR较常规序列差异有统计学意义（P＜0.05；图3）。

点击查看表格

表3

不同AF客观评价以及主观评分的差异性比较

Tab. 3

Comparison of differences in objective evaluation and subjective scores of different AFs

表3

不同AF客观评价以及主观评分的差异性比较

Tab. 3

Comparison of differences in objective evaluation and subjective scores of different AFs

	常规	SENSE 2	CS 2	CS 3	CS 4	CS 5	F值	P值
矢状位T1WI
椎体SNR	45.45±8.44	43.00±9.92	46.26±10.05	42.30±9.02	40.72±8.75	33.66±6.38	8.365	＜0.001
椎体CNR	31.41±8.08	29.86±9.03	32.42±9.39	29.46±8.36	28.54±8.07	23.76±6.21	4.304	0.001
间盘SNR	25.89±3.73	24.32±3.78	26.51±4.63	24.34±3.26	23.35±3.48	19.33±2.95	15.290	＜0.001
间盘CNR	11.85±3.13	11.18±3.09	12.67±3.63	11.54±2.86	11.17±3.00	9.45±2.47	3.921	0.002
评分	4.88±0.34	4.88±0.34	4.84±0.37	4.78±0.42	4.75±0.44	3.28±0.63	67.682	＜0.001
矢状位T2WI
椎体SNR	17.99±5.59	16.63±5.55	18.16±6.39	16.48±4.96	14.57±4.79	13.17±4.77	4.252	0.001
椎体CNR	26.21±7.40	24.38±7.84	25.22±9.05	24.59±7.28	22.06±7.01	19.54±5.93	3.412	0.006
间盘SNR	10.98±3.71	10.10±3.51	11.11±4.42	10.17±3.49	9.04±3.31	8.12±2.72	3.344	0.006
间盘CNR	33.22±8.49	30.92±8.55	32.27±10.44	30.90±8.25	27.59±7.46	24.59±7.76	4.579	0.001
评分	4.88±0.34	4.81±0.40	4.81±0.40	4.78±0.42	4.34±0.70	3.00±0.72	63.552	＜0.001
轴位T2WI
椎体SNR	20.74±4.46	19.36±4.28	22.05±6.33	18.30±4.36	10.83±3.12	8.10±1.78	56.741	＜0.001
椎体CNR	16.37±4.63	15.63±4.80	17.93±4.63	14.45±5.06	7.96±2.86	5.26±1.62	48.841	＜0.001
间盘SNR	20.93±7.77	19.73±8.03	22.30±8.80	18.77±8.29	11.10±5.54	8.34±4.52	19.514	＜0.001
间盘CNR	16.56±8.26	16.16±8.26	18.19±8.10	15.21±8.25	8.40±5.21	5.67±4.36	15.602	＜0.001
评分	4.91±0.30	4.84±0.37	4.84±0.37	4.75±0.44	4.34±0.70	2.03±0.70	158.642	＜0.001

注：SENSE为敏感度编码；CS为压缩感知；2、3、4、5分别为加速因子值；SNR为信噪比；CNR为对比噪声比。

点击查看大图

图3

不同加速序列椎体信噪比（SNR）、椎体对比噪声比（CNR）、间盘SNR、间盘CNR箱式图。3A～3D：矢状位T1WI序列；3E～3H：矢状位T2WI序列；3I～3L：轴位T2WI序列。当压缩感知（CS）=4时，矢状位T1WI序列椎体和间盘的SNR、矢状位T2WI序列椎体和间盘的SNR、CNR与常规序列差异有统计学意义；当CS=3时，轴位T2WI序列椎体SNR与常规序列差异有统计学意义。SENSE为敏感度编码。

Fig. 3

Box diagram of signal to noise ratio (SNR) of vertebral body, contrast to noise ratio (CNR) of vertebral body, SNR of intervertebral disc and CNR of intervertebral disc in different acceleration sequence. 3A-3D: T1WI sagittal positions; 3E-3H: T2WI sagittal positions; 3I-3L: T2WI axial positions. When compressed sensing (CS)=4, SNR of sagittal T1WI vertebral body and intervertebral disc, SNR and CNR of sagittal T2WI vertebral body and intervertebral disc are significantly different from those of conventional sequences; When CS=3, the SNR of vertebral body on axial T2WI is statistically different from that of conventional sequence. SENSE is sensitivity encoding.

点击查看大图

图3

Fig. 3

3 讨论

本研究将CS应用于2D腰椎矢状位T1WI、T2WI和轴位T2WI三个序列，优化AF。结果表明当AF为3、3、2时，椎体SNR、CNR和主观评分与常规序列差异无统计学意义，扫描时间缩短56.65%，扫描效率提高一倍以上，相关研究国内尚未有报道。通过CS使腰椎MRI时间显著缩短，不仅可以提高检查成功率，还可提高设备的社会经济效益，具有较好的临床应用前景。

3.1 腰椎MRI检查常规序列的优势及局限性

随着年龄的增长，腰椎退行性病变的发病率逐渐增高^{[14, 15]}，常出现慢性腰背疼、下肢麻木疼痛等症状，严重者引发压缩骨折^[16]。计算机断层成像（computed tomography, CT）可以较好显示腰椎形态特征，广泛用于脊柱外科手术计划和术后随访。但CT对椎间盘的显示较差，且电离辐射无法避免^{[17, 18]}。MRI因具有良好的软组织对比度、较高空间分辨率、无创伤、无电离辐射等优势，通过多序列、多方位扫描，用以评估椎体及间盘的形态学改变，被认为是腰椎及间盘病变的首选影像学检测^{[19, 20]}。腰椎MRI检查是临床最常见的检查部位之一，但因其检查时间相对较长^[21]，对于新发压缩骨折、严重脊椎后凸的患者，因其难以长时间保持仰卧位姿势，常会因不自主运动导致图像质量产生伪影。在保证图像质量的前提下缩短扫描时间，不仅可以显著提高检查成功率，还可以有效提升MR设备的社会经济效益。

3.2 CS原理及应用优势

CS是由信号稀疏分解和逼近理论进一步发展形成的一种信号处理技术^{[22, 23]}。依据MR信号具有稀疏性和不相干性的特点，通过数字化随机稀疏采样的方式获得离散样本的信号^[24]，然后采用非线性重建算法迭代重建获得稀疏信号，实现了部分K空间数据采集成像^{[25, 26]}。由于采样频率远小于奈奎斯特采样定律，扫描时间明显缩短^{[27, 28]}。该技术最先由LUSTIG等^[29]应用于临床。目前CS的临床应用逐渐推广，各部位不同序列最佳AF优化成为了国内外研究热点^{[30, 31]}。浦仁旺等^[32]将CS应用于3D屏气胰胆管成像，对比传统序列，当AF为24时，图像质量仍能满足临床诊断需求，扫描时间仅为15 s。林青等^[33]通过比较心脏MRI CS序列与常规序列，结果表明采用CS序列的图像与常规序列图像无明显差异，心功能分析准确可靠，且检查时间明显缩短。王学东等^[34]利用结合CS的mDixon Quant序列进行肝脏脂肪定量，定量结果不受影响，且扫描时间明显缩短。

3.3 CS应用于腰椎2D成像的优势及可行性

常规腰椎MRI检查包括矢状位T1WI、T2WI和轴位T2WI三个序列，一次检查时间约10 min，已成为临床最常见的检查之一。但对于部分受检者，因慢性腰背疼等原因难以忍受长时间检查，缩短扫描时间可以有效提高检查成功率。BRATKE等^[35]将CS与腰椎3D-T2WI序列结合，结果表明当应用AF=4.5时，扫描时间缩短39%，图像质量与应用SENSE的序列无明显差异。MORITA等^[12]将CS应用于腰椎3D-T2WI-VISTA序列，可以在保证图像质量的前提下明显缩短扫描时间。然而，由于3D序列扫描时间相对较长，2D序列在临床的应用更加广泛。所以本研究将CS与腰椎2D MRI序列相结合，探究CS对2D腰椎MRI图像的影响，以推广其临床应用价值。BRATKE等^[13]将CS应用于腰椎矢状位2D TSE-T2WI与GRE-T2WI序列扫描，结果表明TSE序列较GRE序列伪影减少，且可以使用更高的AF。对于TSE序列，AF为3时图像质量满足诊断需求，扫描时间缩短43%。本研究根据前者的研究，创新性地将CS应用于腰椎矢状位T1WI、T2WI和轴位T2WI三个序列。当AF为3时，矢状位T2WI序列椎体、间盘SNR、CNR和主观评分与常规序列无明显差异，扫描时间缩短65.83%（103.0 s vs. 35.2 s），与前者研究结果一致，但节省了更多的扫描时间。对于矢状位T1WI和轴位T2WI序列，AF分别为3和2时，扫描时间缩短64.01%（162.0 vs. 58.3 s）、46.76%（216.0 vs. 115.0 s），且对陈旧性椎体压缩骨折、许莫氏结节、腰间盘突出、椎体脂肪化等腰椎退行性病变显示良好，图像质量满足诊断需求。

3.4 本研究的局限性

本研究存在以下局限性：（1）样本量较小，有待进一步扩大样本量；（2）椎体退行性病变因患者不同年龄、性别而异^[36]，因患者量较小，未将患者年龄、性别等因素纳入考虑范围内；（3）因扫描时间较长，将怀疑新发压缩骨折患者、严重脊椎后凸受检者排除，但后续临床扫描中显示，应用CS的2D序列对以上病变的显示良好，且大部分患者能够配合完成检查。

4 结论

综上所述，对比传统序列，结合CS的腰椎2D序列可以在保证图像质量的前提下，有效缩短扫描时间，提高检查成功率。临床推荐AF分别为3、3、2行腰椎矢状位T1WI、T2WI和轴位T2WI序列扫描，对腰间盘突出、陈旧性椎体压缩骨折等退行性病变显示清晰，部分新发压缩骨折患者、严重脊椎后凸患者也可配合完成检查。且总扫描时间缩短56.65%，扫描效率提高一倍以上，可以显著提高MR设备的社会经济效益，减少患者预约等待时间，具有一定的推广应用价值。

本文引用格式：

张浩南, 宋清伟, 张楠, 等. 压缩感知技术在腰椎磁共振快速成像中的应用[J]. 磁共振成像, 2023, 14(2): 132-137, 144.

Cite this article as:

ZHANG H N, SONG Q W, ZHANG N, et al. Application of compressed sensing technology in rapid lumbar magnetic resonance imaging[J]. Chin J Magn Reson Imaging, 2023, 14(2): 132-137, 144.

利益冲突

作者利益冲突声明：

全体作者均声明无利益冲突。

参考文献References

[1]

LINF G, ZHOUX Z, ZHANGB, et al. Utility of flexion-extension radiographs with brackets and magnetic resonance facet fluid for the assessment of lumbar instability in Degenerative Lumbar Spondylolisthesis[J/OL]. World Neurosurg, 2022: S1878-S8750(22)01225-6 [2022-10-28]. https://www.sciencedirect.com/science/article/abs/pii/S1878875022012256. DOI: 10.1016/j.wneu.2022.08.115.

[2]

XUC Q, YINM C, MOW. Correlation and differences in lumbopelvic sagittal alignment parameters between lumbar radiographs and magnetic resonance images[J]. Global Spine J, 2022, 12(1): 79-84. DOI: 10.1177/2192568220947049.

[3]

WUP H, HUANGC Y, LIW H, et al. Application of magnetic resonance diffusion tensor imaging in the clinical diagnosis of disc herniation after lumbar spine injury[J/OL]. J Healthc Eng, 2021, 2021: 6610988 [2022-09-28]. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7969098/. DOI: 10.1155/2021/6610988.

[4]

张旭胤, 郭银霞, MichaelPAXON Julien, 等. T2 mapping定量评价腰椎间盘退变的Meta分析[J]. 磁共振成像, 2022, 13(8): 55-59, 70. DOI: 10.12015/issn.1674-8034.2022.08.010.

ZHANGX Y, GUOY X, PAXONJ M, et al. The value of quantitative assessment of lumbar disc degeneration with T2 mapping: a Meta analysis[J]. Chin J Magn Reson Imaging, 2022, 13(8): 55-59, 70. DOI: 10.12015/issn.1674-8034.2022.08.010.

[5]

LORENCT, GOŁĘBIOWSKIM, MICHALSKIW, et al. High-resolution, three-dimensional magnetic resonance imaging axial load dynamic study improves diagnostics of the lumbar spine in clinical practice[J]. World J Orthop, 2022, 13(1): 87-101. DOI: 10.5312/wjo.v13.i1.87.

[6]

张月青, 丁金立, 龚涛, 等. 基于压缩感知3D-FLAIR序列与常规2D-FLAIR序列在脑白质病变成像中的对比研究[J]. 磁共振成像, 2022, 13(4): 100-102, 119. DOI: 10.12015/issn.1674-8034.2022.04.018.

ZHANGY Q, DINGJ L, GONGT, et al. Comparative study of 3D-FLAIR sequence based on compressed sensing and the conventional 2D-FLAIR sequence in white matter lesion imaging[J]. Chin J Magn Reson Imaging, 2022, 13(4): 100-102, 119. DOI: 10.12015/issn.1674-8034.2022.04.018.

[7]

夏爽, 祁吉. MR并行采集技术的优势[J]. 国外医学(临床放射学分册), 2006, 29(5): 348-353. DOI: 10.3784/j.issn.1674-1897.2006.05.016.

XIAS, QIJ. Advantages of MR parallel acquisition technology[J]. Foreign Med Sci Clin Radiol Fascicle, 2006, 29(5): 348-353. DOI: 10.3784/j.issn.1674-1897.2006.05.016.

[8]

CROPF, GUILLAUDO, BEN HAJ AMORM, et al. Comparison of compressed sensing and controlled aliasing in parallel imaging acceleration for 3D magnetic resonance imaging for radiotherapy preparation[J]. Phys Imaging Radiat Oncol, 2022, 23: 44-47. DOI: 10.1016/j.phro.2022.06.008.

[9]

VORNETTIG, BARTIROMOF, TONIF, et al. Follow-up assessment of intracranial aneurysms treated with endovascular coiling: comparison of compressed sensing and parallel imaging time-of-flight magnetic resonance angiography[J]. Tomography, 2022, 8(3): 1608-1617. DOI: 10.3390/tomography8030133.

[10]

KIDOT, HIRAIK, OGAWAR, et al. Comparison between conventional and compressed sensing cine cardiovascular magnetic resonance for feature tracking global circumferential strain assessment[J/OL]. J Cardiovasc Magn Reson, 2021, 23(1): 10 [2022-12-29]. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7898736/. DOI: 10.1186/s12968-021-00708-5.

[11]

NARESH NK, MALONEL, FUJIWARAT, et al. Use of compressed sensing to reduce scan time and breath-holding for cardiac cine balanced steady-state free precession magnetic resonance imaging in children and young adults[J]. Pediatr Radiol, 2021, 51(7): 1192-1201. DOI: 10.1007/s00247-020-04952-2.

[12]

MORITAK, NAKAURAT, MARUYAMAN, et al. Hybrid of compressed sensing and parallel imaging applied to three-dimensional isotropic T₂-weighted turbo spin-echo MR imaging of the lumbar spine[J]. Magn Reson Med Sci, 2020, 19(1): 48-55. DOI: 10.2463/mrms.mp.2018-0132.

[13]

BRATKEG, RAUR, KABBASCHC, et al. Speeding up the clinical routine: compressed sensing for 2D imaging of lumbar spine disc herniation[J/OL]. Eur J Radiol, 2021, 140: 109738 [2022-10-25]. https://www.ejradiology.com/article/S0720-048X(21)00219-9/fulltext. DOI: 10.1016/j.ejrad.2021.109738.

[14]

HUANG RF, PAN FM, KONGC, et al. Age- and sex-dependent differences in the morphology and composition of paraspinal muscles between subjects with and without lumbar degenerative diseases[J/OL]. BMC Musculoskelet Disord, 2022, 23(1): 734 [2022-12-29]. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9341069/. DOI: 10.1186/s12891-022-05692-0.

[15]

HIDAT, EASTLACK RK, KANEMURAT, et al. Effect of race, age, and gender on lumbar muscle volume and fat infiltration in the degenerative spine[J]. J Orthop Sci, 2021, 26(1): 69-74. DOI: 10.1016/j.jos.2019.09.006.

[16]

GAOF, LIUS, ZHANG XD, et al. Automated grading of lumbar disc degeneration using a push-pull regularization network based on MRI[J]. J Magn Reson Imaging, 2021, 53(3): 799-806. DOI: 10.1002/jmri.27400.

[17]

FUKUSHIMAY, MATSUDAA, KOIZUMIK, et al. Usefulness of model-based iterative reconstruction in low-dose lumbar spine CT after spine surgery with metal implant: a phantom study[J]. Acta Radiol, 2021, 62(10): 1333-1340. DOI: 10.1177/0284185120961424.

[18]

MORBÉEL, CHENM, HERREGODSN, et al. MRI-based synthetic CT of the lumbar spine: geometric measurements for surgery planning in comparison with CT[J/OL]. Eur J Radiol, 2021, 144: 109999 [2022-10-13]. https://www.ejradiology.com/article/S0720-048X(21)00480-0/fulltext. DOI: 10.1016/j.ejrad.2021.109999.

[19]

SHAYGANFARA, KHODAYIM, EBRAHIMIANS, et al. Quantitative diagnosis of osteoporosis using lumbar spine signal intensity in magnetic resonance imaging[J/OL]. Br J Radiol, 2019, 92(1097): 20180774 [2022-10-13]. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6580909. DOI: 10.1259/bjr.20180774.

[20]

HUANG CC, JAW FS, YOUNG YH. Radiological and functional assessment in patients with lumbar spinal stenosis[J/OL]. BMC Musculoskelet Disord, 2022, 23(1): 137 [2022-12-29]. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8829986/. DOI: 10.1186/s12891-022-05053-x.

[21]

DE BUCK M HS, JEZZARDP, HESS AT. Optimization of undersampling parameters for 3D intracranial compressed sensing MR angiography at 7 T[J]. Magn Reson Med, 2022, 88(2): 880-889. DOI: 10.1002/mrm.29236.

[22]

GARREAUM, PUISEUXT, TOUPINS, et al. Accelerated sequences of 4D flow MRI using GRAPPA and compressed sensing: a comparison against conventional MRI and computational fluid dynamics[J]. Magn Reson Med, 2022, 88(6): 2432-2446. DOI: 10.1002/mrm.29404.

[23]

OBAMAY, OHNOY, YAMAMOTOK, et al. MR imaging for shoulder diseases: effect of compressed sensing and deep learning reconstruction on examination time and imaging quality compared with that of parallel imaging[J]. Magn Reson Imaging, 2022, 94: 56-63. DOI: 10.1016/j.mri.2022.08.004.

[24]

AOIKET, FUJIMAN, YONEYAMAM, et al. Development of three-dimensional MR neurography using an optimized combination of compressed sensing and parallel imaging[J]. Magn Reson Imaging, 2022, 87: 32-37. DOI: 10.1016/j.mri.2021.12.002.

[25]

KIMD, HEO YJ, JEONG HW, et al. Compressed sensing time-of-flight magnetic resonance angiography with high spatial resolution for evaluating intracranial aneurysms: comparison with digital subtraction angiography[J]. Neuroradiol J, 2021, 34(3): 213-221. DOI: 10.1177/1971400920988099.

[26]

BENJAMIN A JV, BANOW, MAIRG, et al. Diagnostic quality assessment of IR-prepared 3D magnetic resonance neuroimaging accelerated using compressed sensing and k-space sampling order optimization[J]. Magn Reson Imaging, 2020, 74: 31-45. DOI: 10.1016/j.mri.2020.08.025.

[27]

KIMJ, SEON, BAEH, et al. Comparison of sensitivity encoding (SENSE) and compressed sensing-SENSE for contrast-enhanced T1-weighted imaging in patients with crohn disease undergoing MR enterography[J]. AJR Am J Roentgenol, 2022, 218(4): 678-686. DOI: 10.2214/AJR.21.26733.

[28]

REN SJ, WUW, SU CQ, et al. High-resolution compressed sensing time-of-flight MR angiography outperforms CT angiography for evaluating patients with Moyamoya disease after surgical revascularization[J/OL]. BMC Med Imaging, 2022, 22(1): 64 [2022-12-29]. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC898840/. DOI: 10.1186/s12880-022-00790-w.

[29]

LUSTIGM, DONOHOD, PAULY JM. Sparse MRI: the application of compressed sensing for rapid MR imaging[J]. Magn Reson Med, 2007, 58(6): 1182-1195. DOI: 10.1002/mrm.21391.

[30]

CHOI ES, KIM JS, NICKEL MD, et al. Free-breathing contrast-enhanced multiphase MRI of the liver in patients with a high risk of breath-holding failure: comparison of compressed sensing-accelerated radial and Cartesian acquisition techniques[J]. Acta Radiol, 2022, 63(11): 1453-1462. DOI: 10.1177/02841851211052988.

[31]

YAMADAT, MASUIT, SASAKIM, et al. Time resolved DCE-MRI of the kidneys: evaluation of the renal vasculatures and tumors using F-DISCO with and without compressed sensing in normal and wide-bore 3T systems[J/OL]. Medicine (Baltimore), 2022, 101(31): e29971 [2022-09-28]. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9351894.

[32]

浦仁旺, 刘爱连, 王家正, 等. 不同压缩感知加速倍数对三维磁共振胰胆管成像图像质量的影响[J]. 实用放射学杂志, 2021, 37(1): 128-131. DOI: 10.3969/j.issn.1002-1671.2021.01.031.

PU RW, LIU AL, WANG JZ, et al. Influence of different compressed-sensing acceleration factors on three-dimensional magnetic resonance cholangiopancreatography image quality[J]. J Pract Radiol, 2021, 37(1): 128-131. DOI: 10.3969/j.issn.1002-1671.2021.01.031.

[33]

林青, 王佳佳, 孙明华, 等. 心脏磁共振压缩感知电影序列的临床应用[J]. 中国医学影像技术, 2020, 36(2): 281-286. DOI: 10.13929/j.issn.1003-3289.2020.02.027.

LINQ, WANG JJ, SUN MH, et al. Clinical application of cardiac magnetic resonance compressed sensing cine imaging[J]. Chin J Med Imaging Technol, 2020, 36(2): 281-286. DOI: 10.13929/j.issn.1003-3289.2020.02.027.

[34]

王学东, 刘爱连, 张钦和, 等. 压缩感知技术对3D mDIXON Quant定量分析肝脏脂肪的影响[J]. 中国医学影像技术, 2020, 36(11): 1662-1666. DOI: 10.13929/j.issn.1003-3289.2020.11.016.

WANG XD, LIU AL, ZHANG QH, et al. Impact of compressed sensing technology on 3D mDIXON Quant liver fat quantification[J]. Chin J Med Imaging Technol, 2020, 36(11): 1662-1666. DOI: 10.13929/j.issn.1003-3289.2020.11.016.

[35]

BRATKEG, RAUR, WEISSK, et al. Accelerated MRI of the lumbar spine using compressed sensing: quality and efficiency[J/OL]. J Magn Reson Imaging, 2019, 49(7): e164-e175 [2022-09-28]. https://onlinelibrary.wiley.com/doi/10.1002/jmri.26526. DOI: 10.1002/jmri.26526.

[36]

宋宇, 宋清伟, 张浩南, 等. 基于压缩感知的3DmDixonQuant 对腰椎椎体脂肪含量的定量研究[J]. 磁共振成像, 2021, 12(4): 51-56. DOI: 10.12015/issn.1674-8034.2021.04.010.

SONGY, SONG QW, ZHANG HN, et al. 3D mDixon Quant based on compressed SENSE for quantitative study of lumbar vertebral body fat content[J]. Chin J Magn Reson Imaging, 2021, 12(4): 51-56. DOI: 10.12015/issn.1674-8034.2021.04.010.

贡献者信息

张浩南

大连医科大学附属第一医院放射科，大连　116011

宋清伟

大连医科大学附属第一医院放射科，大连　116011

张楠

大连医科大学附属第一医院放射科，大连　116011

宋宇

四川大学华西第二医院放射科，成都　610041

浦仁旺

大连医科大学附属第一医院放射科，大连　116011

王楠

大连医科大学附属第一医院放射科，大连　116011

刘爱连

大连医科大学附属第一医院放射科，大连　116011

通信作者

宋清伟

大连医科大学附属第一医院放射科，大连　116011

Email：songqw1964@163.com

关键词

腰椎; 磁共振成像; 压缩感知; 二维; 运动伪影;

基金项目

大连医科大学附属第一医院“压缩感知在全身的应用价值”项目（2021HZ006）

作者声明

作者贡献声明：

张浩南起草和撰写稿件，获取、分析或解释本研究的数据；宋清伟设计本研究的方案，解释本研究的数据，对稿件重要内容进行了修改，获得大连医科大学附属第一医院“压缩感知在全身的应用价值”项目基金支持；张楠、宋宇、浦仁旺、王楠、刘爱连获取、分析本研究的数据，对稿件重要内容进行了修改；全体作者都同意发表最后的修改稿，同意对本研究的所有方面负责，确保本研究的准确性和诚信。

利益冲突

作者利益冲突声明：

全体作者均声明无利益冲突。

历史

出版日期：2023-02-20

收稿日期：2022-10-06

录用日期：2023-01-17

Technical Article

Application of compressed sensing technology in rapid lumbar magnetic resonance imaging

ZHANG Haonan, SONG Qingwei, ZHANG Nan, SONG Yu, PU Renwang, WANG Nan, LIU Ailian

Published 2023-02-20

Cite as Chin J Magn Reson Imaging, 2023, 14(2): 132-137,144. DOI: 10.12015/issn.1674-8034.2023.02.022

Abstract

Objective

To explore the influence of different acceleration factors (AF) of compressed sensing (CS) on the quality of lumbar MRI images.

Materials and Methods

Thirty-two subjects (twelve males) were recruited with an average age of (45.28±14.11) years. The sagittal T1WI, T2WI, and axial T2WI sequences of the lumbar spine were scanned with 3.0 T MR equipment through the sensitivity encoding (SENSE), and CS technology for AF=0, SENSE AF=2, CS AF=2, 3, 4, and 5, respectively. Two radiologists delineated the region of interest (ROI) on the sagittal T1WI, T2WI and transverse T2WI to measure the signal intensity (SI) and standard deviation (SD). Then we calculated the signal to noise ratio (SNR) and contrast to noise ratio (CNR). Finally, subjective score of image quality was assessed by five points method. The intra-class correlation coefficient (ICC) and Kappa test was adopted to evaluate the consistency of the scores from the two radiologists. In the following analysis, the ANOVA test was used to assess the difference of SNR, CNR and score between groups.

Results

The measured datas and the subjective score of the two radiologists were in good agreement (ICC: 0.878-0.997, Kappa: 0.763-0.948). It was shown that there were statistically significant differences in SNR, CNR and subjective score of sagittal T1WI, T2WI and transverse T2WI sequences. If the CS equaled 4, the SNR, CNR and subjective score of sagittal T1WI and T2WI were significantly different from those of conventional sequences (P＜0.05). If the CS equaled 3, the SNR, CNR and subjective score of axial T2WI were significantly different from conventional sequences (P＜0.05).

Conclusions

Scan time for the lumbar spine decreased gradually with increase of the CS AF. CS factor of 3 was recommended for clinical sagittal T1WI and T2WI, and CS factor of 2 was best for clinical transverse T2WI to achieve an optimal balance between scan time and image quality.

Key words:

lumbar spine; magnetic resonance imaging; compressed sensing; two dimensional; motion artifact

Contributor Information

ZHANG Haonan