分享:
分享到微信朋友圈
X
技术研究
头颈联合3D-TOF-MRA人工智能辅助压缩感知序列的优化
袁畅 张煜堃 曹家骏 宋清伟 苗延巍

Cite this article as: YUAN C, ZHANG Y K, CAO J J, et al. Optimization of artificial intelligence-assisted compressed sensing sequences for cerebral and carotid 3D-TOF-MRA[J]. Chin J Magn Reson Imaging, 2024, 15(4): 139-144, 152.本文引用格式:袁畅, 张煜堃, 曹家骏, 等. 头颈联合3D-TOF-MRA人工智能辅助压缩感知序列的优化[J]. 磁共振成像, 2024, 15(4): 139-144, 152. DOI:10.12015/issn.1674-8034.2024.04.022.


[摘要] 目的 通过比较施加不同加速因子(acceleration factors, AF)的并行采集(parallel imaging, PI)、压缩感知(compressed sensing, CS)、人工智能-压缩感知(artificial intelligence compressed sensing, ACS)技术的头颈联合三维时间飞跃法磁共振血管成像(three-dimensional time-of-fight magnetic resonance angiography, 3D-TOF-MRA)的图像质量,选取优化序列。材料与方法 前瞻性招募24例健康志愿者,以PI的AF为3(PI 3)、CS的AF分别为4和6(CS 4、CS 6)以及ACS的AF分别为4、6、8、10(ACS 4、ACS 6、ACS 8、ACS 10)进行头颈联合3D-TOF-MRA扫描。扫描时间:PI 3=8 min 40 s;CS 4=6 min 38 s;CS 6=4 min 9 s;ACS 4=5 min 24 s;ACS 6=4 min 30 s;ACS 8=4 min 13 s;ACS 10=3 min 24 s。分别在双侧大脑中动脉(middle cerebral artery, MCA)M1段、双侧颈内动脉(internal carotid artery, ICA)C4段、颈动脉分叉处下5个层面的双侧颈总动脉(common carotid artery, CCA)以及MCA及ICA同一层面的颞叶白质,CCA同一层面的胸锁乳突肌作为背景区域勾画感兴趣区(regions of interest, ROI),记录信号强度(signal intensity, SI)及标准差(standard deviation, SD),从而计算信噪比(signal-to-noise ratio, SNR)与对比噪声比(contrast-to-noise ratio, CNR)。采用四分法和五分法分别对整体图像质量以及颅内动脉、颈部大动脉进行评分。以组内相关系数(intra-class correlation coefficient, ICC)比较两名放射医师之间及同一名放射医师内客观评价结果的一致性,Kappa检验比较两名放射医师之间及同一名放射医师内主观评价结果的一致性,若一致性良好,则选取其中一位医师的客观评分及主观评分进行后续统计分析。采用单因素方差分析或Kruskal-Wallis检验对客观评分以及主观评分进行总体差异比较,若差异具有统计学意义则进行两两比较。结果 与PI技术比较显示,ACS 4~ACS 8的SNRL-MCA,ACS 8、ACS 10的CNRL-MCA,ACS 4、ACS 6、ACS 10的CNRR-MCA,ACS 4~ACS 10的SNRR-MCA、SNRL-CCA、CNRL-CCA,ACS 6~ACS 10的SNRR-CCA、CNRR-CCA均高于PI 3,差异具有统计学意义(P<0.05)。与CS技术比较显示,ACS 4~ACS 10的SNRR-MCA、CNRL-MCA及CNRR-MCA,ACS 4~ACS 8的SNRL-MCA、SNRL-CCA、SNRR-CCA、CNRL-CCA、CNRR-CCA与CS 4差异具有统计学意义(P<0.05),ACS 4~ACS 10的SNRL-MCA、SNRR-MCA、CNRL-MCA、CNRR-MCA、SNRL-ICA、SNRR-ICA、CNRL-ICA、SNRL-CCA、SNRR-CCA、CNRL-CCA、CNRR-CCA与CS 6差异具有统计学意义(P<0.05),均优于CS 4、CS 6。ACS技术之间两两比较,ACS 8的SNRL-MCA高于ACS 10(P<0.05),ACS 8、ACS 10的SNRR-CCA、CNRR-CCA均高于ACS 4(P<0.05),其余差异均无统计学意义(P>0.05)。除颈部大动脉图像外,主观评分统计结果均为ACS 4~ACS 10与CS 6差异具有统计学意义(P<0.05),优于CS 6。结论 与PI及CS技术相比,ACS技术拥有更短的扫描时间,更好的图像质量。ACS 8为最优序列,扫描时间比PI 3缩短51%。
[Abstract] Objective By comparing the application of different acceleration factors (AF) in parallel imaging (PI), compressed sensing (CS), and artificial intelligence-compressed sensing (ACS) techniques for three-dimensional time-of-flight magnetic resonance angiography (3D-TOF-MRA) in the cerebral and carotid region, to select an optimized acceleration factor for ACS.Materials and Methods Twenty-four healthy volunteers were prospectively recruited and underwent cerebral and carotid 3D-TOF-MRA scanning with AF of 3 in PI (PI 3), 4 and 6 in CS (CS 4 and CS 6), and 4, 6, 8 and 10 in ACS (ACS 4, ACS 6, ACS 8 and ACS 10). Scan durations were: PI 3=8 min 40 s; CS 4=6 min 38 s; CS 6=4 min 9 s; ACS 4=5 min 24 s; ACS 6=4 min 30 s; ACS 8=4 min 13 s; ACS 10=3 min 24 s. Regions of interest (ROIs) were delineated in bilateral middle cerebral artery (MCA) at M1 segment, bilateral internal carotid artery (ICA) at C4 segment, five levels below the bifurcation of the carotid artery in bilateral common carotid artery (CCA), and temporal white matter at the same level as MCA and ICA, with the sternocleidomastoid muscle at the same level as CCA serving as the background region. Signal intensity (SI) and standard deviation (SD) were recorded to calculate signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR). Image quality was assessed using four-point and five-point scales for overall image quality, intracranial arteries, and major neck arteries. Intra-class correlation coefficient (ICC) was used to compare the consistency of objective evaluations between two radiologists and within the same radiologist, while Kappa test was employed to compare the consistency of subjective evaluations. If consistency was satisfactory, objective and subjective ratings from one radiologist were selected for subsequent analysis. One-way ANOVA or Kruskal-Wallis test was used to compare overall differences in objective and subjective ratings, with post hoc tests conducted for statistically significant differences.Results Compared with PI, there were statistically significant differences (P<0.05) in SNRL-MCA of ACS 4-ACS 8, CNRL-MCA of ACS 8 and ACS 10, CNRR-MCA of ACS 4, ACS 6, and ACS 10, SNRR-MCA, SNRL-CCA, CNRL-CCA of ACS 4-ACS 10, SNRR-CCA, CNRR-CCA, compared with PI 3, all of which were higher than PI 3. Compared with CS, statistically significant differences (P<0.05) were observed in SNRR-MCA, CNRL-MCA, and CNRR-MCA of ACS 4-ACS 10, SNRL-MCA, SNRL-CCA, SNRR-CCA, CNRL-CCA, CNRR-CCA of ACS 4-ACS 8, compared with CS 4, as well as SNRL-MCA, SNRR-MCA, CNRL-MCA, CNRR-MCA, SNRL-ICA, SNRR-ICA, CNRL-ICA, SNRL-CCA, SNRR-CCA, CNRL-CCA, CNRR-CCA of ACS 4-ACS 10, compared with CS 6, all of which were superior to CS 4 and CS 6. In pairwise comparisons among ACS, SNRL-MCA of ACS 8 was higher than ACS 10 (P<0.05), SNRR-CCA and CNRR-CCA of ACS 8 and ACS 10 were both higher than ACS4 (P<0.05), while other differences were not statistically significant (P>0.05). Except for the images of the carotid arteries, subjective scoring results showed statistically significant differences (P<0.05) between ACS 4-ACS 10 and CS 6, all of which were superior to CS 6.Conclusions Compared with PI and CS, ACS has shorter scanning time and better image quality. ACS 8 is the optimal sequence, with a 51% reduction in scanning time compared with PI 3.
[关键词] 并行采集;压缩感知;人工智能-压缩感知;头颈联合三维时间飞跃法磁共振血管成像;磁共振成像
[Keywords] parallel imaging;compressed sensing;artificial intelligence-compressed sensing;cerebral and carotid combined time-of-flight magnetic resonance angiography;magnetic resonance imaging

袁畅    张煜堃    曹家骏    宋清伟    苗延巍 *  

大连医科大学附属第一医院放射科,大连 116011

通信作者:苗延巍,E-mail:ywmiao716@163.com

作者贡献声明:袁畅起草和撰写稿件,获取、分析并解释本研究数据;苗延巍设计本研究的方案,对稿件重要内容进行修改;张煜堃、曹家骏、宋清伟分析或解释本研究数据,对稿件重要内容进行修改;宋清伟获得辽宁省教育厅科学研究经费项目资助。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。


基金项目: 辽宁省教育厅科学研究经费项目 LJKZ0856
收稿日期:2023-11-12
接受日期:2024-03-22
中图分类号:R445.2  R743.3 
文献标识码:A
DOI: 10.12015/issn.1674-8034.2024.04.022
本文引用格式:袁畅, 张煜堃, 曹家骏, 等. 头颈联合3D-TOF-MRA人工智能辅助压缩感知序列的优化[J]. 磁共振成像, 2024, 15(4): 139-144, 152. DOI:10.12015/issn.1674-8034.2024.04.022.

0 引言

       颅内外动脉粥样硬化性狭窄及闭塞是缺血性卒中的重要病因[1, 2],准确评估血管狭窄程度对于该疾病的治疗和患者的预后具有重要意义[3, 4, 5]。数字减影血管造影(digital subtraction angiography, DSA)是评价头颈动脉狭窄的金标准,但有创性限制了其在临床中的广泛应用[6, 7]。为此,非侵入性的血管成像方法,如计算机断层扫描血管造影(computed tomography angiography, CTA)和头颈联合三维时间飞跃法磁共振血管成像(three-dimensional time-of-flight magnetic resonance angiography, 3D-TOF-MRA)成为备受关注的替代选择。CTA通过静脉注入碘对比剂显示颈动脉狭窄,但碘对比剂有过敏风险并可能对肾脏功能产生影响[8]。磁共振3D-TOF-MRA检查是一种无创性血管成像方法,避免了对比剂的使用和辐射暴露,具有较好的安全性和患者接受性,可用于头部和颈部血管的评估[9, 10]。头颈部联合3D-TOF-MRA提供了更丰富和完整的颅内外动脉的诊断信息,临床应用日益增多。然而,由于头颈部联合3D-TOF-MRA的扫描范围较大,导致扫描时间相对较长,限制了其在临床中的应用。近年来磁共振成像加速技术的发展为缩短扫描时间提供了新的方法,包括并行采集(parallel imaging, PI)[11]及压缩感知(compressed sensing, CS)[12]技术等,它们通过改进数据采集和图像重建过程,实现了扫描时间的显著缩短,以应用于头颈部血管磁共振成像(magnetic resonance angiography, MRA)[13, 14]。最近,人工智能(artificial intelligence, AI)的引入为加速图像处理过程提供了新的可能性。AI辅助CS(artificial intelligence compressed sensing, ACS)技术结合了CS和AI,旨在保证图像质量的前提下进一步缩短扫描时间[15]。然而,目前仅存在CS技术加速3D-TOF-MRA应用于头颅血管的报道[16],尚缺乏PI、CS、ACS同时应用于头颈部动脉3D-TOF-MRA的详尽报道。因此,本研究扩大扫描范围,拟通过对比施加不同加速因子(acceleration factors, AF)的PI、CS和ACS的头颈部联合3D-TOF-MRA图像质量,寻找扫描时间与图像质量最优搭配的序列。从而在短时间内提供更丰富的头颈部动脉影像学信息,并提高患者磁共振成像检查的舒适性,减少运动伪影。

1 材料与方法

1.1 一般资料

       本研究于2023年6月至8月在大连医科大学附属第一医院招募24例健康志愿者,进行头颈联合3D-TOF-MRA检查。其中男8例,女16例,年龄21~69岁,中位年龄38.5岁。纳入标准:(1)成年人,年龄20~70岁之间;(2)能配合完成全部序列检查。排除标准:(1)磁共振成像检查禁忌证;(2)中枢神经系统及心血管、代谢等疾病;(3)图像存在伪影。本研究遵守《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,批文号:PJ-KS-KY-2022-274,全体受试者均签署了知情同意书。

1.2 仪器与扫描方法

       采用联影Omega 3.0 T磁共振扫描仪,48通道头颈联合相控阵线圈采集3D-TOF-MRA图像。志愿者取仰卧位,头先进,身体与床体保持一致,双手置于身体两侧。3D-TOF-MRA采用AF为3的PI技术(PI 3)[17];AF分别为4、6的CS技术(CS 4、CS 6);AF分别为4、6、8、10的ACS技术(ACS 4、ACS 6、ACS 8、ACS 10)进行扫描。除了AF不同以外,各序列的其他扫描参数均相同:TR 20.1 ms,TE 4.4 ms,FOV 220 mm×200 mm,翻转角18°,体素大小0.60 mm×0.60 mm×1.60 mm。扫描时间:PI 3=8 min 40 s;CS 4=6 min 38 s;CS 6=4 min 9 s;ACS 4=5 min 24 s;ACS 6=4 min 30 s;ACS 8=4 min 13 s;ACS 10=3 min 24 s。

1.3 图像质量评价

1.3.1 客观评价

       由一名具有5年诊断经验的主治医师(医师1)以及一名具有10年诊断经验的副主任医师(医师2)完成图像主客观评分,间隔1个月由医师2再次重复测量。在头颈联合3D-TOF-MRA的原始图像上分别于双侧大脑中动脉(middle cerebral artery, MCA)M1段、双侧颈内动脉(internal carotid artery, ICA)C4段、颈动脉分叉处下5个层面的双侧颈总动脉(common carotid artery, CCA)以及MCA及ICA同一层面的颞叶白质,CCA同一层面的胸锁乳突肌作为背景区域勾画感兴趣区(region of interest, ROI),血管处勾画大小为10~15 mm²的ROI,背景区域处勾画大小为30 mm²的ROI,见图1。测量各ROI信号强度(signal intensity, SI)和标准差(standard deviation, SD),计算信噪比(signal-to-noise ratio, SNR),对比度噪声比(contrast-to-noise ratio, CNR)[16],见公式(1)~(2)。

图1  三维时间飞跃法磁共振血管成像(3D-TOF-MRA)原始图像中感兴趣区(ROI)勾画示意图。1A:大脑中动脉;1B:颈内动脉;1C:颈总动脉。
Fig. 1  Schematic diagram of region of interest (ROI) delineation in raw three-dimensional time-of-flight magnetic resonance angiography (3D-TOF-MRA) images. 1A: Middle cerebral artery; 1B: Internal carotid artery; 1C: Common carotid artery.

1.3.2 主观评价

       将图像导入工作站,对头颈联合3D-TOF-MRA以最大密度投影(maximal intensity projection, MIP)进行重建。由上述医师1和医师2以四分法[18]对整体图像质量进行评价:4分,图像优异,血管边缘显示锐利,血流信号均匀无流动伪影;3分,图像良好,管腔显示良好,可有轻微不光滑,血管信号均匀,有轻微流动伪影;2分,图像可接受,管壁欠光滑,血流信号不均匀;1分,图像无法诊断,血管未显示。以五分制量表[19]分别对双侧颈总动脉(common carotid artery, LR-CCA)、双侧颈内动脉(internal carotid artery, LR-ICA)、双侧椎动脉(vertebral artery, LR-VA)、基底动脉(basilar artery, BA)双侧大脑前动脉(anterior cerebral artery, LR-ACA)、双侧大脑中动脉(middle cerebral artery, LR-MCA)、双侧大脑后动脉(posterior cerebral artery, LR-PCA)进行评价:4分,血管连续明显可视;3分,血管连续较弱可视;2分,血管具有小间隙,非连续可视;1分,血管具有较大间隙,非连续可视;0分,非可视。

1.4 统计学分析

       采用SPSS 25.0统计分析软件。为评估两名放射医师之间及医师2两次测量之间客观评分和主观评分结果的一致性,使用组内相关系数(intraclass correlation coefficient, ICC)和Kappa检验,即ICC/Kappa≤0.20为一致性极差,0.20<ICC/Kappa≤0.40为一致性较差,0.40<ICC/Kappa≤0.60为一致性一般,0.60<ICC/Kappa≤0.80为一致性较好,而ICC/Kappa>0.80则表示一致性好。采用Shapiro-Wilkin对数据进行正态性分析。采用单因素方差分析或Kruskal-Wallis检验对客观评分和主观评分进行总体差异比较,若差异具有统计学意义,将进行组间两两比较。符合正态性的数据以均数±标准差表示;不符合正态性的数据以中位数(上下四分位数)表示。P<0.05为差异具有统计学意义。

2 结果

2.1 一致性检验

       一致性检验结果显示,两名放射医师之间图像客观评分ICC为0.827~0.887,图像主观评分Kappa为0.700~0.942,一致性较好(P<0.05)。医师2两次测量结果客观评分ICC为0.876~0.953,图像主观评分Kappa为0.872~0.979,一致性好(P<0.05)。以医师2第一次测量的客观评分及主观评分结果进行统计学分析(表1)。

表1  两名医师客观评分及主观评分一致性比较
Tab. 1  Comparison of objective and subjective scores between the two physicians

2.2 客观评价

       除CNRR-ICA外,各血管原始图像SNR及CNR的总体比较差异均具有统计学意义(P均<0.05)。与PI图像相比,在双侧大脑中动脉中,ACS 4~ACS 8的SNRL-MCA,ACS 4~ACS 10的SNRR-MCA,ACS 8、ACS 10的CNRL-MCA,ACS 4、ACS 6、ACS 10的CNRR-MCA与PI 3差异具有统计学意义(P<0.05),均高于PI 3;在双侧颈内动脉中,各参数间差异均无统计学意义(P>0.05);在双侧颈总动脉中,ACS 4~ACS 10的SNRL-CCA、CNRL-CCA,ACS 6~ACS 10的SNRR-CCA、CNRR-CCA与PI 3差异具有统计学意义(P<0.05)。而与CS图像相比,在双侧大脑中动脉中,ACS 4~ACS 10的SNRR-MCA、CNRL-MCA及CNRR-MCA,ACS 4~ACS 8的SNRL-MCA与CS4差异具有统计学意义(P<0.05),ACS 4~ACS 10的SNRL-MCA、SNRR-MCA、CNRL-MCA及CNRR-MCA与CS 6差异具有统计学意义(P<0.05);在双侧颈内动脉中ACS 4~ACS 10的SNRL-ICA、SNRR-ICA及CNRL-ICA与CS 6差异具有统计学意义(P<0.05);在双侧颈总动脉中,差异均具有统计学意义(P<0.05),均高于CS 4、CS 6。ACS之间两两比较显示,ACS 8的SNRL-MCA高于ACS 10(P<0.05),ACS 8、ACS 10的SNRR-CCA、CNRR-CCA均高于ACS 4(P<0.05),其余差异均无统计学意义(P>0.05),见表2, 3, 4

表2  3D-TOF-MRA大脑中动脉客观评价分析
Tab. 2  Analysis of objective evaluation of middle cerebral artery by 3D-TOF-MRA
表3  3D-TOF-MRA颈内动脉客观评价分析
Tab. 3  Analysis of objective evaluation of internal carotid artery by 3D-TOF-MRA
表4  3D-TOF-MRA颈总动脉客观评价分析
Tab. 4  Analysis of objective evaluation of common carotid arteries by 3D-TOF-MRA

2.3 主观评价

       7组不同AF的颅内动脉图像以及图像总体质量的主观评分总体差异均具有统计学意义(P<0.05),颈部大动脉图像质量的主观评分总体差异不具有统计学意义(P>0.05);颅内动脉图像以及整体图像质量两两比较结果均为CS 6与ACS 4、ACS 6、ACS 8、ACS 10差异具有统计学意义(P<0.05),具体见表5图2

图2  女,23岁,健康志愿者。冠状位头颈联合3D-TOF-MRA MIP图像。2A~2G分别为PI 3、CS 4、CS 6、ACS 4、ACS 6、ACS 8、ACS 10图像,图像整体评分分别为4、3、3、4、4、4、4分。3D-TOF-MRA:三维时间飞跃法磁共振血管成像;PI:并行采集;CS:压缩感知;ACS:人工智能辅助压缩感知。
Fig. 2  Female, 23 years old, healthy volunteer. Coronal 3D-TOF-MRA MIP images of the cerebral and carotid junction. 2A-2G represent PI 3, CS 4, CS 6, ACS 4, ACS 6, ACS 8, ACS 10 images, with overall image scores of 4, 3, 3, 4, 4, 4, and 4, respectively. 3D-TOF-MRA: three-dimensional time-of-flight magnetic resonance angiography; PI: parallel imaging; CS: compressed sensing; ACS: artificial intelligence-assisted compressed sensing.
表5  3D-TOF-MRA主观评分
Tab. 5  Subjective scoring for 3D-TOF-MRA

3 讨论

       由于受到时间限制,3D-TOF-MRA的扫描范围较小。头颈联合3D-TOF-MRA扩大了扫描范围,可以同时显示头颈动脉血管,并且与DSA相比具有良好的一致性[20]。然而,头颈联合3D-TOF-MRA扫描时间长,使患者难以配合,加大扫描风险和图像伪影的出现概率。本研究首次将ACS技术应用于头颈联合3D-TOF-MRA中提高扫描速度,并与PI技术和CS技术进行了全面比较。结果发现,在客观评价中ACS技术大部分图像的SNR、CNR高于PI和CS,差异具有统计学意义。ACS 8的SNRL-MCA显著高于ACS 10,ACS 8、ACS 10的SNRR-CCA、CNRR-CCA均显著高于ACS 4;主观评价中,ACS 4~ACS 10的整体图像质量、颅内动脉图像评分均显著高于CS 6。因此,3D-TOF-MRA ACS 8为最优序列。扫描时间的减少不仅能够减缓患者的不适感和焦虑情绪,还能提高医院资源利用效率,在医疗资源优化利用和经济效益提升方面显现出积极的前景。

3.1 与现有基于PI技术及CS技术的3D-TOF-MRA相比

       PI技术是一种通过增加相控阵线圈的数量来减少扫描时间的方法。然而,PI技术随着AF的增大,SNR逐渐下降,并常会出现重建伪影及卷褶伪影等问题[21, 22]。近年来,CS技术广泛用于颅脑磁共振检查中,并且相较于PI技术,CS技术在颅脑及颈部3D-TOF MRA中具有显著的优势,包括提高图像采集速度、降低图像伪影,更好的可视性[14, 23]。REN等[24]的研究发现,在评估手术血运重建后的烟雾病患者中,CS TOF-MRA相较于CTA表现更为优越,能更清晰地显示新生血管。另一项研究[25]以DSA为诊断标准,证实了CS TOF-MRA作为检查头颈部动脉狭窄闭塞性疾病的可靠性。本研究亦发现,CS 4和CS 6在与PI 3的主客观评分比较中差异均无统计学意义,并且扫描时间更短。然而,CS技术虽然缩短了扫描时间,但其客观测量值均小于PI 3。这可以解释为CS技术通过对K空间进行大量且随机的欠采样,并利用迭代算法来寻找最优解来缩短扫描时间,而当CS的AF较大时,迭代重建效率受到影响,从而降低了图像质量[26]

3.2 基于ACS技术的3D-TOF-MRA的临床价值及可行性

       ACS技术对CS技术进行了优化,结合了CS、PI和半傅立叶采集技术,并引入基于深度学习神经网络的AI模块,该神经网络以不同AF下取得的欠采样数据作为输入,以相应的全采样数据作为金标准进行训练。将压缩的AI模块集成到CS框架的迭代重建程序中,实现加速扫描并获得更好的图像质量[27, 28, 29]。近几年,ACS逐渐应用于踝关节,膝关节磁共振成像中,ZHAO等[30]的研究表明,与PI及CS相比,ACS可显著加速踝关节扫描速度,将T1、T2和质子加权序列的采集时间缩短了32%~43%,同时保持了诊断图像的高质量。此外,在三维磁共振序列的应用中,ACS有助于优化膝关节磁共振成像,显示出代替传统CS技术的巨大的潜力[31, 32]。本研究同样对PI、CS、ACS三种加速技术在头颈联合3D-TOF-MRA中的应用进行了比较。结果显示,ACS 4、ACS 6、ACS 8的SNR和CNR均显著高于与PI 3,而与CS技术相比,ACS 4、ACS 6、ACS 8、ACS 10的SNR、CNR以及图像主观评分均显著优于CS技术所对应的图像。ACS 8具有相对清晰的头颈部血管的可视化和较短的扫描时间,相对于PI 3缩短了51%,可作为临床常规扫描的一种实用设置。

3.3 本研究的局限性

       本研究也具有一定局限性:(1)只对健康志愿者进行扫描,未纳入脑血管疾病患者,对于疾病患者的适用性有待考察;(2)样本量相对少,存在一定的偶然性,需要进一步开展基于多中心前瞻性研究,扩大样本量;(3)手动勾画ROI,可能具有一定偏差;(4)ACS技术具有一定不稳定性,我们发现随着ACS技术AF的增大,部分血管的SNR、CNR反而逐渐增加,这可能主要由于数据的规模将影响ACS技术的准确性[33],其次无法完全避免由于采样不足或稀疏度不足而导致的混叠伪像[34]

4 结论

       综上所述,通过对比PI、CS和ACS技术在头颈部3D-TOF-MRA中的应用,本研究发现ACS技术在客观评价和主观评价中表现更出色,具有较高的图像质量和较短的扫描时间。此外,在选择不同ACS AF时,ACS 8在SNR、CNR、扫描时间等方面显示出相对较好的性能。因此,推荐ACS技术在头颈部3D-TOF-MRA中的应用,特别是在ACS 8 AF下,可在4 min 13 s内获得良好的图像质量。这为磁共振血管成像技术的进一步发展提供了新的思路和方向。

[1]
PU Y H, ZOU X Y, WANG Y L, et al. Difference of one year death and stroke recurrence in ischemic stroke patients with anterior and posterior circulation intracranial atherosclerosis[J]. Zhonghua Yi Xue Za Zhi, 2018, 98(7): 502-507. DOI: 10.3760/cma.j.issn.0376-2491.2018.07.004.
[2]
ZHAO D L, LI R Y, LI C, et al. Assessment of the degree of arterial stenosis in intracranial atherosclerosis using 3D high-resolution MRI: comparison with time-of-flight MRA, contrast-enhanced MRA, and DSA[J]. Clin Radiol, 2023, 78(2): e63-e70. DOI: 10.1016/j.crad.2022.08.132.
[3]
NIU J X, RAN Y C, CHEN R, et al. Use of PETRA-MRA to assess intracranial arterial stenosis: comparison with TOF-MRA, CTA, and DSA[J/OL]. Front Neurol, 2022, 13: 1068132 [2023-11-27]. https://pubmed.ncbi.nlm.nih.gov/36726752/. DOI: 10.3389/fneur.2022.1068132.
[4]
MUNIO M, DARCOURT J, GOLLION C, et al. Large artery intracranial stenosis in young adults with ischaemic stroke[J]. Rev Neurol, 2022, 178(3): 206-212. DOI: 10.1016/j.neurol.2021.10.008.
[5]
GWAK D S, KIM B K, CHUNG I, et al. The usefulness of Time-of-Flight MR angiography in detection of intraplaque hemorrhage in patients with acute ischemic stroke with symptomatic carotid stenosis[J/OL]. PLoS One, 2020, 15(2): e0229024 [2023-11-27]. https://pubmed.ncbi.nlm.nih.gov/32053703/. DOI: 10.1371/journal.pone.0229024.
[6]
YU J F, PUNG L, MINAMI H, et al. Virtual 2D angiography from four-dimensional digital subtraction angiography (4D-DSA): a feasibility study[J]. Interv Neuroradiol, 2021, 27(2): 307-313. DOI: 10.1177/1591019920961604.
[7]
HUANG T, ZOU S L, SHEN X, et al. Evaluation of the level of carotid bifurcation and the morphology of extracranial internal carotid artery in patients with carotid stenosis by color Doppler ultrasound and digital subtraction angiography[J]. Zhonghua Yi Xue Za Zhi, 2022, 102(11): 781-786. DOI: 10.3760/cma.j.cn112137-20210605-01285.
[8]
SAMARZIJA K, MILOSEVIC P, JURJEVIC Z, et al. Grading of carotid artery stenosis with computed tomography angiography: whether to use the narrowest diameter or the cross-sectional area[J]. Insights Imaging, 2018, 9(4): 527-534. DOI: 10.1007/s13244-018-0622-5.
[9]
WU C H, CHEN S T, CHEN J H, et al. Diagnosis of extracranial carotid stenosis by MRA of the brain[J/OL]. Sci Rep, 2021, 11(1): 12010 [2023-11-27]. https://pubmed.ncbi.nlm.nih.gov/34103593/. DOI: 10.1038/s41598-021-91511-w.
[10]
YOKUŞ A, TOPRAK N, GÜNDÜZ A M, et al. Anterior cerebral artery and anterior communicating artery variations: assessment with magnetic resonance angiography[J/OL]. World Neurosurg, 2021, 155: e203-e209 [2023-11-27]. https://pubmed.ncbi.nlm.nih.gov/34400324/. DOI: 10.1016/j.wneu.2021.08.027.
[11]
FUJITA H. New horizons in MR technology: RF coil designs and trends[J]. Magn Reson Med Sci, 2007, 6(1): 29-42. DOI: 10.2463/mrms.6.29.
[12]
ARUN A, THOMAS T J, RANI J S, et al. Efficient directionality-driven dictionary learning for compressive sensing magnetic resonance imaging reconstruction[J/OL]. J Med Imaging, 2020, 7(1): 014002 [2023-11-27]. https://pubmed.ncbi.nlm.nih.gov/32042856/. DOI: 10.1117/1.JMI.7.1.014002.
[13]
SAKATA A, FUSHIMI Y, OKADA T, et al. Evaluation of cerebral arteriovenous shunts: a comparison of parallel imaging time-of-flight magnetic resonance angiography (TOF-MRA) and compressed sensing TOF-MRA to digital subtraction angiography[J]. Neuroradiology, 2021, 63(6): 879-887. DOI: 10.1007/s00234-020-02581-y.
[14]
张浩南, 宋清伟, 王家正, 等. 初探压缩感知高倍数加速因子对颈动脉3D-TOF-MRA的影响[J]. 临床放射学杂志, 2021, 40(6): 1079-1083. DOI: 10.13437/j.cnki.jcr.2021.06.010.
ZHANG H N, SONG Q W, WANG J Z, et al. A preliminary study on the effect of compressed sensing with multiple acceleration factors on carotid 3D-TOF-MRA[J]. J Clin Radiol, 2021, 40(6): 1079-1083. DOI: 10.13437/j.cnki.jcr.2021.06.010.
[15]
黄超, 胡军武, 李丽, 等. 人工智能辅助压缩感知技术用于腰椎MRI[J]. 中国医学影像技术, 2023, 39(7): 1080-1083. DOI: 10.13929/j.issn.1003-3289.2023.07.029.
HUANG C, HU J W, LI L, et al. Artificial intelligence-assisted compressed sensing technology for lumbar MR scanning[J]. Chin J Med Imag Technol, 2023, 39(7): 1080-1083. DOI: 10.13929/j.issn.1003-3289.2023.07.029.
[16]
陈楚玥, 于雨洁, 王茂雪, 等. 基于人工智能重建对颅内3D TOF MRA图像质量的改善[J]. 中国CT和MRI杂志, 2023, 21(10): 60-63. DOI: 10.3969/j.issn.1672-5131.2023.10.019.
CHEN C Y, YU Y J, WANG M X, et al. Improvement of intracranial 3D TOF MRA image quality based on artificial intelligence reconstruction[J]. Chin J CT MRI, 2023, 21(10): 60-63. DOI: 10.3969/j.issn.1672-5131.2023.10.019.
[17]
李贝贝, 张浩南, 方鑫, 等. 不同压缩感知加速因子对头部3D-T1WI成像质量及基于体素的形态测量定量参数的影响[J]. 中国医学影像技术, 2022, 38(11): 1730-1734. DOI: 10.13929/j.issn.1003-3289.2022.11.031.
LI B B, ZHANG H N, FANG X, et al. Impacts of compressed sensing acceleration factors on imaging quality of head 3D-T1WI and voxel-based morphometry quantitative parameters[J]. Chin J Med Imag Technol, 2022, 38(11): 1730-1734. DOI: 10.13929/j.issn.1003-3289.2022.11.031.
[18]
钮恩斌, 陈海伦, 周雪梅, 等. 压缩感知并行采集在头颅磁共振血管成像中的应用[J]. 实用医学影像杂志, 2023, 24(2): 90-93. DOI: 10.16106/j.cnki.cn14-1281/r.2023.02.002.
NIU E B, CHEN H L, ZHOU X M, et al. Application of compressed sensing combined with parallel imaging in intracranial MRA[J]. J Pract Med Imag, 2023, 24(2): 90-93. DOI: 10.16106/j.cnki.cn14-1281/r.2023.02.002.
[19]
张煜堃, 刘娜, 曹家骏, 等. 基于敏感性编码及压缩感知技术优化颅脑MR三维相位对比静脉造影成像序列[J]. 中国医学影像技术, 2023, 39(6): 915-919. DOI: 10.13929/j.issn.1003-3289.2023.06.026.
ZHANG Y K, LIU N, CAO J J, et al. Optimization of MR sequences of craniocerebral three-dimensional phase-contrast magnetic resonance venogram based on sensitivity coding and compressed sensing technology[J]. Chin J Med Imag Technol, 2023, 39(6): 915-919. DOI: 10.13929/j.issn.1003-3289.2023.06.026.
[20]
WANG Q, WANG G, SUN Q, et al. Application of MAGnetic resonance imaging compilation in acute ischemic stroke[J]. World J Clin Cases, 2021, 9(35): 10828-10837. DOI: 10.12998/wjcc.v9.i35.10828.
[21]
韩林梅, 俞琴. 并行采集技术在1.5 T 8通道相控阵线圈头颅3D TOF MRA中的优化: 体模研究[J]. 中国医学影像技术, 2019, 35(3): 423-427. DOI: 10.13929/j.1003-3289.201807146.
HAN L M, YU Q. Optimization of parallel acquisition technique in 3DTOF MRA of head with 8-receiving channels phased-array coil in1.5TMR system: a phantom study[J]. Chin J Med Imag Technol, 2019, 35(3): 423-427. DOI: 10.13929/j.1003-3289.201807146.
[22]
泮智勇, 丁雪委, 王世威, 等. 32通道头部线圈与并行采集技术iPAT在头颅MRA检查中的应用[J]. 医学影像学杂志, 2013, 23(11): 1673-1675, 1682.
PAN Z Y, DING X W, WANG S W, et al. Application of 32-channel head coil with integrated parallel acquisition technique(iPAT)in the skull of MRA[J]. J Med Imag, 2013, 23(11): 1673-1675, 1682.
[23]
DING J, DUAN Y, ZHUO Z, et al. Acceleration of brain TOF-MRA with compressed sensitivity encoding: a multicenter clinical study[J]. AJNR Am J Neuroradiol, 2021, 42(7): 1208-1215. DOI: 10.3174/ajnr.A7091.
[24]
REN S J, WU W, SU C Q, et al. High-resolution compressed sensing time-of-flight MR angiography outperforms CT angiography for evaluating patients with Moyamoya disease after surgical revascularization[J]. BMC Med Imaging, 2022, 22(1): 64. DOI: 10.1186/s12880-022-00790-w.
[25]
ZHANG X, CAO Y Z, MU X H, et al. Highly accelerated compressed sensing time-of-flight magnetic resonance angiography may be reliable for diagnosing head and neck arterial steno-occlusive disease: a comparative study with digital subtraction angiography[J]. Eur Radiol, 2020, 30(6): 3059-3065. DOI: 10.1007/s00330-020-06682-3.
[26]
GAO T Y, LU Z, WANG F Z, et al. Using the compressed sensing technique for lumbar vertebrae imaging: comparison with conventional parallel imaging[J]. Curr Med Imaging, 2021, 17(8): 1010-1017. DOI: 10.2174/1573405617666210126155814.
[27]
WANG Q Z, ZHAO W L, XING X Y, et al. Feasibility of AI-assisted compressed sensing protocols in knee MR imaging: a prospective multi-reader study[J]. Eur Radiol, 2023, 33(12): 8585-8596. DOI: 10.1007/s00330-023-09823-6.
[28]
DRATSCH T, SIEDEK F, ZÄSKE C, et al. Reconstruction of shoulder MRI using deep learning and compressed sensing: a validation study on healthy volunteers[J/OL]. Eur Radiol Exp, 2023, 7(1): 66 [2023-11-27]. https://pubmed.ncbi.nlm.nih.gov/37880546/. DOI: 10.1186/s41747-023-00377-2.
[29]
FERVERS P, ZAESKE C, RAUEN P, et al. Conventional and deep-learning-based image reconstructions of undersampled K-space data of the lumbar spine using compressed sensing in MRI: a comparative study on 20 subjects[J/OL]. Diagnostics, 2023, 13(3): 418 [2023-11-27]. https://pubmed.ncbi.nlm.nih.gov/36766523/. DOI: 10.3390/diagnostics13030418.
[30]
ZHAO Q, XU J J, YANG Y X, et al. AI-assisted accelerated MRI of the ankle: clinical practice assessment[J/OL]. Eur Radiol Exp, 2023, 7(1): 62 [2023-11-27]. https://pubmed.ncbi.nlm.nih.gov/37857868/. DOI: 10.1186/s41747-023-00374-5.
[31]
NI M, HE M, YANG Y X, et al. Application research of AI-assisted compressed sensing technology in MRI scanning of the knee joint: 3D-MRI perspective[J/OL]. Eur Radiol, 2023 [2023-11-27]. https://pubmed.ncbi.nlm.nih.gov/37932390/. DOI: 10.1007/s00330-023-10368-x.
[32]
IUGA A I, RAUEN P S, SIEDEK F, et al. A deep learning-based reconstruction approach for accelerated magnetic resonance image of the knee with compressed sense: evaluation in healthy volunteers[J/OL]. Br J Radiol, 2023, 96(1146): 20220074 [2023-11-27]. https://pubmed.ncbi.nlm.nih.gov/37086077/. DOI: 10.1259/bjr.20220074.
[33]
钟美梦, 曹家骏, 安琪, 等. 人工智能辅助压缩感知技术对颅脑磁敏感加权成像图像质量及核团相位值的影响[J]. 磁共振成像, 2023, 14(12): 91-97. DOI: 10.12015/issn.1674-8034.2023.12.015.
ZHONG M M, CAO J J, AN Q, et al. Impact of AI-assisted compressed sensing on quality and phase value of nuclei of brain susceptibility weighted imaging[J]. Chin J Magn Reson Imag, 2023, 14(12): 91-97. DOI: 10.12015/issn.1674-8034.2023.12.015.
[34]
SUI H, GONG Y, LIU L, et al. Comparison of artificial intelligence-assisted compressed sensing (ACS) and routine two-dimensional sequences on lumbar spine imaging[J/OL]. J Pain Res, 2023, 16: 257-267 [2023-11-27]. https://pubmed.ncbi.nlm.nih.gov/36744117/. DOI: 10.2147/JPR.S388219.

上一篇 基于深度学习重建技术的头部增强T1WI序列在垂体神经内分泌肿瘤病变成像中的应用
下一篇 基于模态交互学习的多源心脏图像分割方法研究
  
诚聘英才 | 广告合作 | 免责声明 | 版权声明
联系电话:010-67113815
京ICP备19028836号-2