分享:
分享到微信朋友圈
X
技术研究
人工智能-压缩感知技术在颅脑3D T2-FLAIR序列采集及脑白质高信号评价中的应用
曹家骏 刘娜 钟美梦 袁畅 张煜堃 苗延巍 宋清伟

Cite this article as: CAO J J, LIU N, ZHONG M M, et al. Application of artificial intelligence-assisted compressed sensing technology in brain 3D T2-FLAIR sequence acquisition and evaluation of white matter hyperintensity[J]. Chin J Magn Reson Imaging, 2024, 15(2): 135-139, 146.本文引用格式曹家骏, 刘娜, 钟美梦, 等. 人工智能-压缩感知技术在颅脑3D T2-FLAIR序列采集及脑白质高信号评价中的应用[J]. 磁共振成像, 2024, 15(2): 135-139, 146. DOI:10.12015/issn.1674-8034.2024.02.020.


[摘要] 目的 探究不同的基于人工智能压缩感知(artificial intelligence-assisted compressed sensing, ACS)加速因子对颅脑3D T2WI液体衰减反转恢复(3D T2WI fluid-attenuated inversion-recovery, 3D T2-FLAIR)序列图像质量的影响,并获取最优化的扫描方案。材料与方法 前瞻性纳入健康青年志愿者(healthy control, HC)25例、脑白质高信号(white matter hyperintensity, WMH)患者15例,HC组分别以并行采集(parallel imaging, PI)技术(加速因子为3,F3)和不同加速因子(3、4、5、6、7、8)ACS技术采集颅脑3D T2-FLAIR图像,测量双侧半卵圆中心、双侧尾状核、胼胝体压部、双侧红核、双侧黑质、脑桥、双侧小脑的信号强度以及标准差,并计算图像的信噪比(signal to noise ratio, SNR)和对比噪声比(contrast to noise ratio, CNR)。采用五分法对图像质量进行主观评分。采用组内相关系数(intra-class correlation coefficient, ICC)、Kappa检验比较前后两次测量及观察者间主观评分的一致性。对不同加速因子的图像的SNR、CNR及主观评分采用Friedman秩和检验进行对照分析,综合评判后得出最佳的ACS加速因子;WMH组分别以F3及最佳ACS加速因子采集颅脑3D T2-FLAIR图像,并由两名经验丰富的诊断医师对脑白质病灶数目、Fazekas分级进行评估,采用独立样本t检验、Mann-Whitney U检验进行对照分析。结果 HC组中,不同3D T2-FLAIR的SNR、CNR及主观评分差异具有统计学意义(P均<0.05);两两比较结果显示,3D T2-FLAIRACS3、3D T2-FLAIRACS4与3D T2-FLAIRF3的SNR、CNR,3D T2-FLAIRACS3、3D T2-FLAIRACS4、3D T2-FLAIRACS5与3D T2-FLAIRF3的主观评分差异无统计学意义(P均>0.05),其余图像SNR、CNR及主观评分差异均具有统计学意义(P均<0.05)。WMH组中,3D T2-FLAIRF3与3D T2-FLAIRACS4在病灶数目和Fazekas分级方面差异无统计学意义(P均>0.05)。结论 以ACS技术采集颅脑3D T2-FLAIR可在保证图像质量和序列诊断效能的前提下缩短扫描时间,选取的最优加速因子为ACS4。
[Abstract] Objective To investigate the effects of different acceleration factors based on artificial intelligence-assisted compressed sensing (ACS) on the image quality of 3D T2WI fluid-attenuated inversion-recovery (3D T2-FLAIR) sequence.Materials and Methods Twenty-five healthy volunteers (HC) and fifteen patients with white matter hyperintensity (WMH) were prospectively included in the study. In HC group, the brain 3D T2-FLAIR images were collected by parallel imaging (PI) technique (parallel acquisition acceleration factor was 3) and ACS technique with different acceleration factors (3, 4, 5, 6, 7, 8). The signal intensity (SI) and standard deviation (SD) of all 3D T2-FLAIR images were measured in bilateral centrum semiovale, bilateral caudate nucleus, splenium of corpus callosum, bilateral red nucleus, bilateral substantia nigra, pons and bilateral cerebellum. The signal to noise ratio (SNR) and contrast to noise ratio (CNR) were further calculated. The subjective score of image quality was analyzed according to five grades standard. The intra-class correlation coefficient (ICC) and Kappa test were used to compare the consistency between the measured data and the subjective scores of the two observers. The SNR, CNR and subjective scores of images with different acceleration factors were compared by Friedman test. After comprehensive evaluation, the best ACS acceleration factor is obtained. In the WMH group, 3D T2-FLAIR images of the brain were collected with F3 and the best ACS acceleration factor, and the number of WMH and Fazekas grades were evaluated by two experienced diagnostic physicians. Independent sample t test and Mann-Whitney U test were used for comparative analysis.Results In HC group, The SNR, CNR and subjective scores of different 3D T2-FLAIR sequences were statistically significant (all P<0.05). The results of pairwise comparison showed that the SNR and CNR of 3D T2-FLAIRACS3, 3D T2-FLAIRACS4 and 3D T2-FLAIRF3, and the subjective scores of 3D T2-FLAIRACS3, 3D T2-FLAIRACS4, 3D T2-FLAIRACS5 and 3D T2-FLAIRF3 were not statistically significant (all P>0.05). The SNR, CNR and subjective scores of the remaining images were statistically significant (all P<0.05). In the WMH group, there was no significant difference in the number of WMH and Fazekas grades between 3D T2-FLAIR F3 and 3D T2-FLAIR ACS4 ( P>0.05 ).Conclusions The acquisition of brain 3D T2-FLAIR with ACS technology can shorten the scanning time under the premise of ensuring image quality and diagnostic efficiency, and ACS4 can be considered as the best acceleration factor.
[关键词] 人工智能-压缩感知;压缩感知;磁共振成像;脑;加速采集
[Keywords] artificial intelligence compressed sensing;compressed sensing;magnetic resonance imaging;brain;acceleration

曹家骏    刘娜    钟美梦    袁畅    张煜堃    苗延巍    宋清伟 *  

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,E-mail:songqw1964@163.com

作者贡献声明::曹家骏起草和撰写稿件,获取、分析和解释本研究数据;宋清伟获得辽宁省教育厅科学研究经费项目及横向课题基金项目的资助支持并设计本研究的方案,对稿件重要内容进行修改;刘娜、钟美梦、袁畅、张煜堃、苗延巍分析、解释本研究数据,对稿件重要内容进行修改;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。


基金项目: 辽宁省教育厅科学研究经费项目 LJKZ0856 横向课题基金项目 2021HZ006
收稿日期:2023-08-10
接受日期:2024-01-31
中图分类号:R445.2 
文献标识码:A
DOI: 10.12015/issn.1674-8034.2024.02.020
本文引用格式曹家骏, 刘娜, 钟美梦, 等. 人工智能-压缩感知技术在颅脑3D T2-FLAIR序列采集及脑白质高信号评价中的应用[J]. 磁共振成像, 2024, 15(2): 135-139, 146. DOI:10.12015/issn.1674-8034.2024.02.020.

0 引言

       T2液体衰减反转恢复(T2 fluid-attenuated inversion- recovery, T2-FLAIR)序列是结合反转准备脉冲和T2加权的水抑制序列[1],其具有对于正常脑脊液信号的高度敏感性和有效抑制性,可以增强颅内病变与正常脑组织的对比,有利于脑梗死、脱髓鞘病变等脑白质病变的检出[2, 3, 4]。3D T2-FLAIR相较于传统的2D采集模式具有明显的优势,包含更丰富的解剖信息[5]以及可以进行各向同性多平面重建等[6],但由于扫描时间的延长,部分年幼、老年、危重等受检者难以耐受[7]

       目前临床上广泛应用的加速采集技术包括并行采集(parallel imaging, PI)[8, 9, 10]与压缩感知(compressed sensing, CS)[11, 12, 13]。联合CS技术进行快速采集的3D T2-FLAIR序列被用于多种疾病的诊断与评估,并已被证明可以在加快成像时间的同时,具有同等甚至更佳的图像质量及诊断效能[14, 15]。人工智能-压缩感知技术(artificial intelligence-assisted compressed sensing, ACS)是一种基于卷积神经网络的深度学习智能CS技术,可以对图像进行超高分辨率重建和降噪,在保证甚至优化图像质量的前提下大幅缩短序列的采集时间[16, 17, 18]。目前,国内外尚未见不同加速因子的ACS技术对3D T2-FLAIR加速效率、图像质量及诊断效能的研究报道。本研究将应用不同ACS加速因子的3D T2-FLAIR图像与传统加速采集技术PI进行对比,旨在保证图像质量和满足临床诊断要求的前提下,探讨加速因子对采集时间和图像质量的影响,筛选出最佳的ACS加速因子。

1 材料与方法

1.1 一般资料

       于2023年3月至5月在大连医科大学附属第一医院前瞻性招募健康青年志愿者(healthy control, HC)25例(男5例,女20例);同期招募脑白质高信号(white matter hyperintensity, WMH)患者15例(男7例,女8例),行颅脑MRI检查。HC组入组标准:(1)年龄>18岁;(2)既往体健,无颅脑疾病史。排除标准:(1)MRI检查绝对禁忌证者;(2)无法完成全部检查序列或图像质量不佳者。WMH组入组标准:(1)年龄>18岁;(2)确诊为脑白质病变者,MR图像上脑白质出现斑点样或斑片样高信号;(3)能够耐受MR检查。排除标准:图像有严重的运动伪影。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 检查方法

       采用联影uMR Omega 3.0 T MR扫描仪(上海联影医疗科技有限公司,中国)及32通道头颅线圈采集颅脑MRI图像。HC组分别以常规的并行采集(parallel imaging, PI)技术(并行采集加速因子3,F3)及ACS技术(加速因子为3、4、5、6、7、8)采集头部冠状位3D T2-FLAIR图像,TR 5 000 ms,TE 464.64 ms,TI 1 550 ms,层厚 1 mm,体素大小1 mm×1 mm×1 mm,获得3D T2-FLAIRF3、3D T2-FLAIRACS3、3D T2-FLAIRACS4、3D T2-FLAIRACS5、3D T2-FLAIRACS6、3D T2-FLAIRACS7及3D T2-FLAIRACS8,图像扫描时间分别为415 s、305 s、275 s、175 s、155 s、135 s、115 s及105 s。WMH组分别以F3及综合评判得出的最佳ACS加速因子采集颅脑3D T2-FLAIR图像。

1.3 图像质量分析

1.3.1 图像客观评价

       采用医学影像处理软件(uWS-MR,上海联影医疗科技有限公司,中国)进行图像分析,对原始图像进行轴位和矢状位重建,分别在重建后的轴位图像的同一层面的双侧半卵圆中心、双侧尾状核、胼胝体压部、双侧红核、双侧黑质、脑桥、双侧小脑信号均匀处放置合适大小的圆形感兴趣区(region of interest, ROI),见图1,测量其信号强度(signal intensity, SI)以及标准差(standard deviation, SD),并对双侧测量的结构取平均值。将双侧半卵圆中心信号强度及标准差的均值记为背景信号(signal intensity of background, SIb)和背景标准差(standard deviation of background, SDb),计算图像信噪比(signal to noise ratio, SNR)(SNR=(SI-SIb)/SD)和对比噪声比(contrast to noise ratio, CNR)(CNR=|SI-SIb|/SD2-SDb2),并于一周后重新测量所有数据,计算重测置信度。

图1  感兴趣区示意图。1A:双侧半卵圆中心;1B:双侧尾状核;1C:胼胝体压部;1D:双侧红核;1E:双侧黑质;1F:脑桥;1G:双侧小脑。
Fig. 1  The picture shows a schematic diagram of region of interest. 1A: Bilateral semioval center; 1B: Bilateral caudate nucleus; 1C: Splenium of corpus callosum; 1D: Bilateral red nucleus; 1E: Bilateral substantia nigra; 1F: Pons; 1G: Bilateral cerebellum.

1.3.2 图像主观评价

       主观评价由两位具有5年以上神经影像学诊断经验的医师采用5级评分法[19]进行分析:1分,灰、白质完全显示不清,组织边缘难以分辨,图像整体伪影及噪声极明显;2分,灰、白质显示不清,组织边缘模糊,图像整体伪影明显,噪声较大;3分,灰、白质显示较清晰,组织边缘尚锐利,图像整体略有伪影,噪声一般;4分,灰、白质显示较清晰,组织边缘锐利,图像整体无伪影,噪声较小;5分,灰、白质及组织边缘显示清晰,图像整体无伪影及噪声。

1.3.3 诊断效能评估

       由上述两位医师对不同加速因子采集得到的3D T2-FLAIR图像上显示的WMH病灶个数和Fazekas分级进行统计,鼓励医师对图像进行多平面重建,结果由两位医师商议后得到。

1.4 统计学分析

       采用SPSS 27.0软件进行统计学分析,正态性检验采用Shapiro-Wilk正态性检验,符合正态分布的计量资料以均数±标准差表示,不符合者以中位数(P25,P75)来表示。HC组图像主观与客观评分的组间比较采用Friedman秩和检验;以Kappa检验比较观察者间图像质量主观评分结果的一致;以组内相关系数(intra-class correlation coefficient, ICC)评价定量指标前后测量的一致性。WMH组病灶个数和Fazekas分级比较采用独立样本t检验及Mann-Whitney U检验。P<0.05为差异有统计学意义。

2 结果

2.1 客观评价结果

       重复测量数据与原数据一致性高(ICC=0.995,P<0.001),选取首次测量数据做统计学分析。经正态性检验,各扫描序列的SNR、CNR均不符合正态分布,故采用中位数(P25,P75)表示。

       3D T2-FLAIRF3、3D T2-FLAIRACS3、3D T2-FLAIRACS4、3D T2-FLAIRACS5、3D T2-FLAIRACS6、3D T2-FLAIRACS7及3D T2-FLAIRACS8的SNR和CNR均具有统计学意义(P<0.001)。序列间两两比较结果显示,3D T2-FLAIRACS3、3D T2-FLAIRACS4的SNR、CNR与3D T2-FLAIRF3相比均无统计学意义(P均>0.05),3D T2-FLAIRACS5、3D T2-FLAIRACS6、3D T2-FLAIRACS7及3D T2-FLAIRACS8的SNR、CNR较3D T2-FLAIRF3降低(P均<0.05)(表1图2)。

图2  不同加速因子的3D T2-FLAIR图像SNR、CNR对比。FLAIR:液体衰减反转恢复;SNR:信噪比;CNR:对比噪声比;ACS:人工智能压缩感知。*表示与F3相比差异具有统计学意义,P<0.05。
Fig. 2  Comparison of SNR and CNR of 3D T2-FLAIR images with different acceleration factors. FLAIR: fluid-attenuated inversion-recovery; SNR: signal to noise ratio; CNR: contrast to noise ratio; ACS: artificial intelligence-assisted compressed sensing. * indicates there is a statistically significant difference compared with F3 group, P<0.05.
表1  3D T2-FLAIR成像质量客观评价和主观评分结果比较(n=25)
Tab. 1  Comparison of objective evaluation and subjective scoring of 3D T2-FLAIR imaging quality (n=25)

2.2 主观评分结果

       2名观察者对3D T2-FLAIRF3、3D T2-FLAIRACS3、3D T2-FLAIRACS4、3D T2-FLAIRACS5、3D T2-FLAIRACS6、3D T2-FLAIRACS7及3D T2-FLAIRACS8图像的主观评分一致性较好(Kappa=0.846、0.856、0.871、0.802、0.918、0.812、0.920,P均<0.05),故选择较高年资观察者的主观评分结果进行分析;7组图像主观评分总体差异具有统计学意义(P<0.001),两两比较结果显示3D T2-FLAIRACS3、3D T2-FLAIRACS4、3D T2-FLAIRACS5的主观评分与3D T2-FLAIRF3相比差异无统计学意义(P>0.05),3D T2-FLAIRACS6、3D T2-FLAIRACS7及3D T2-FLAIRACS8图像主观评分低于3D T2-FLAIRF3(P<0.05);在ACS图像中,随着加速因子的增大,图像的主观评分降低(P<0.05)(表1图3)。

图3  女,27岁,主观评分标准对应图像。3A(A1~A7)、3B(B1~B7)及3C(C1~C7)分别为颅脑轴位、矢状位、冠状位3D T2-FLAIRF3(主观评分为5分)、3D T2-FLAIRACS3(主观评分为5分)、3D T2-FLAIRACS4(主观评分为4分)、3D T2-FLAIRACS5(主观评分为4分)、3D T2-FLAIRACS6(主观评分为3分)、3D T2-FLAIRACS7(主观评分为3分)及3D T2-FLAIRACS8(主观评分为2分)图像。FLAIR:液体衰减反转恢复;ACS:人工智能压缩感知。
Fig. 3  Female, 27 years old, the subjective scoring criteria correspond to the image. 3A (A1-A7), 3B (B1-B7) and 3C (C1-C7) were axial, sagittal and coronal 3D T2-FLAIRF3 (the subjective score is 5), 3D T2-FLAIRACS3 (the subjective score is 5), 3D T2-FLAIRACS4 (the subjective score is 4), 3D T2-FLAIRACS5 (the subjective score is 4), 3D T2-FLAIRACS6 (the subjective score is 3), 3D T2-FLAIRACS7 (the subjective score is 3) and 3D T2-FLAIRACS8 (the subjective score is 2) of brain. FLAIR: fluid-attenuated inversion-recovery; ACS: artificial intelligence-assisted compressed sensing.

2.3 诊断效能结果

       经HC组主观与客观评价结果的综合评估,最终以ACS4为加速因子采集WMH组3D T2-FLAIR图像,并以3D T2-FLAIRF3图像作为对照。

       3D T2-FLAIRACS4与3D T2-FLAIRF3相比,在WMH病灶个数和Fazekas分级方面差异均无统计学意义(P均>0.05)(表2图4)。

图4  女,74岁,脑白质病变,Fazekas 2级。3D T2-FLAIRF3与3D T2-FLAIRACS4轴位侧脑室层面对比图。4A:3D T2-FLAIRF3图像;4B:3D T2-FLAIRACS4图像。FLAIR:液体衰减反转恢复;ACS:人工智能压缩感知。
Fig. 4  Female, 74 years old, white matter lesions, Fazekas grade 2. Comparison of axial 3D T2-FLAIRF3 and 3D T2-FLAIRACS4 images of lateral ventricle level. 4A: 3D T2-FLAIRF3; 4B: 3D T2-FLAIRACS4. FLAIR: fluid-attenuated inversion-recovery; ACS: artificial intelligence-assisted compressed sensing.
表2  3D T2-FLAIRF3与3D T2-FLAIRACS4诊断效能对比(n=15)
Tab. 2  Comparison of diagnostic efficacy between 3D T2-FLAIRF3 and 3D T2-FLAIRACS4 sequences (n=15)

3 讨论

       本研究采用不同的ACS加速因子(ACS3、4、5、6、7、8)对颅脑3D T2-FLAIR序列进行快速成像采集,就图像客观评价指标(SNR、CNR)及主观评价指标(医师主观评分)与目前临床上常用的PI(F3)进行对比。研究结果显示,ACS技术能在保证图像质量的前提下缩短扫描时间,ACS4为最大可应用的加速因子,扫描时间缩短33.7%。此外,我们还在WMH疾病检出(病灶个数)与病情分级(Fazekas分级)方面将所得到的最佳ACS加速因子ACS4与F3进行对比,以评价其诊断效能,最终发现ACS4组与F3组在WMH中诊断效能相当。

3.1 常规3D T2-FLAIR序列的优势及局限性

       T2-FLAIR序列是颅脑MR检查的常规序列,采用容积激发的各向同性3D T2-FLAIR序列相较于传统的2D T2-FLAIR序列能够提供更为丰富的诊断信息[20]。一项针对幕下多发性硬化病灶检出数量的研究[21]发现,相较于T2WI自旋回波序列和2D T2-FLAIR序列, 3D T2-FLAIR可以显示更多的病灶,具有更高的诊断效能;LIU等[22]发现结合超感压缩感知技术的3D T2-FLAIR序列在皮质微梗死检测率、图像质量、空间分辨率方面优于2D T2-FLAIR图像;张月青等[23]的研究报道了3D T2-FLAIR序列可以抑制脑脊液搏动伪影,比2D T2-FLAIR序列更有利于脑白质病变的检出。但由于3D序列的扫描时间较长,部分重症患者因疼痛、意识障碍等原因常易产生不自主运动造成的伪影,从而导致图像质量的降低[24, 25]

3.2 与现有ACS相关研究比较

       ACS技术是基于深度学习模式融合卷积神经网络(convolutional neural networks, CNN)、半傅里叶成像、PI和CS的新型加速技术[26],利用数学建模的方法,ACS技术可以将采样误差转换为CS稀疏约束,使其即使在较高的加速因子下也能从欠采样的K空间数据中有效重建MR图像[27]。此前已有研究表明ACS技术与传统CS相比具有更好的加速性能,对细微解剖结构的显示能力更佳[28]。ZHAO等[29]的研究将ACS技术应用于肾脏快速扫描,发现与传统的T2 NAVI技术相比,除了显著缩短扫描时间外,ACS组图像的SNR和CNR均高于常规NAVI组,即客观图像质量有所提高。SUI等[30]使用单一ACS加速因子对腰椎2D常规序列进行加速采集,结果表明ACS加速序列不仅减少了18.9%的扫描时间,而且在健康受试者和患者中基本上保留了与常规2D序列相同的图像质量,且图像伪影也有所减少。LIU等[31]对因脑部疾病导致不自主头部运动的患者行ACS辅助的T2-FLAIR检查,证明了ACS联合T2-FLAIR在颅脑病变筛查中的可行性,但其研究未对ACS的加速效率和临床可选用的最佳加速因子进行讨论。YANG等[32]使用ACS技术对乳腺3D各向同性高分辨T2WI序列进行加速采集,并与常规2D T2WI图像进行对比,发现ACS缩短成像时间的同时,所获得的图像在病灶的检出率、显示细节、整体图像质量、乳腺病灶诊断信息和乳腺组织勾画等方面明显优于常规序列。UEDA等[33]将ACS应用于女性盆腔疾病患者T2WI的图像采集,发现应用ACS技术可以在提高图像信噪比的同时缩短近50%的扫描时间,与本次研究结果相仿。

       上述研究均证实了ACS的临床适用性,但是加速因子升高在缩短扫描时间的同时必然伴随着图像质量的降低,使得对于合理加速因子的选择成为了密切关注的问题[34]。本次研究结果发现随着加速因子的增大,图像客观评价指标SNR、CNR及主观评分逐渐降低;相比于使用PI的3D T2-FLAIRF3,3D T2-FLAIRACS3、3D T2-FLAIRACS4在客观评价与主观评分方面差异均无统计学意义,且扫描时间分别缩短了26.5%和33.7%;3D T2-FLAIRACS5的主观评分与3D T2-FLAIRF3差异无统计学意义,但是图像SNR、CNR下降,故综合考虑不能作为3D T2-FLAIR序列的合理加速因子选择;3D T2-FLAIRACS6、3D T2-FLAIRACS7、3D T2-FLAIRACS8的客观评价与主观评分均低于3D T2-FLAIRF3,难以满足临床诊断要求,可能导致病变的漏诊。在诊断效能的评估中,我们发现在病灶个数和Fazekas分级方面3D T2-FLAIRACS4与3D T2-FLAIRF3差异无统计学意义,不过3D T2-FLAIRACS4漏检了极少数的细微病变,但这并不影响对病变的整体观察及对病情的总体评估。以上结果表明,ACS技术虽具有较好的加速效率与图像重建性能,但过大的加速因子仍会对图像的对比度和空间分辨率产生影响,故针对具体部位及序列应选取合适的ACS加速因子,以达到扫描时间与图像质量之间的平衡。此外我们还在扫描过程中发现,随着ACS加速因子的增大,图像重建时间不同程度地延长。这可能是因为ACS技术是基于CS稀疏采样的基础上实现的,随着加速倍速的升高,采样点减少,过少的采样点导致计算量增加,图像重建过程更为复杂[35]

3.3 局限性

       本研究的局限性在于:(1)本研究为单中心、单一机型设备研究,样本量较少;(2)没有考虑序列采集后的重建时间对采集速度的影响,后续研究将在纳入重建时间的基础上对图像总采集时间进行更全面的评估,进一步深化对ACS技术加速效能的认识。

4 结论

       综上所述,采集头颅3D T2-FLAIR时,应用ACS4可以在保证图像质量和诊断效能的前提下缩短扫描时间;ACS4可考虑作为最佳加速因子。

[1]
FUJIWARA Y, INOUE Y, KANAMOTO M, et al. The use of combined T2-weighted and FLAIR synthetic magnetic resonance images to improve white matter region contrast: a feasibility study[J]. Radiol Phys Technol, 2019, 12(1): 118-125. DOI: 10.1007/s12194-019-00498-7.
[2]
DONG X, NAO J. Fluid-attenuated inversion recovery vascular hyperintensities in anterior circulation acute ischemic stroke: associations with cortical brain infarct volume and 90-day prognosis[J]. Neurol Sci, 2019, 40: 1675-1682. DOI: 10.1007/s10072-019-03909-0.
[3]
BHARGAVA P, KIM S, REYES A A, et al. Imaging meningeal inflammation in CNS autoimmunity identifies a therapeutic role for BTK inhibition[J]. Brain2021, 144: 1396-1408. DOI: 10.1093/brain/awab045.
[4]
GWO C Y, ZHU D C, ZHANG R. Brain white matter hyperintensity lesion characterization in 3D T2 fluid-attenuated inversion recovery magnetic resonance images: Shape, texture, and their correlations with potential growth[J/OL]. Front Neurosci, 2022, 16: 1028929 [2023-08-10]. https://pubmed.ncbi.nlm.nih.gov/36507337/. DOI: 10.3389/fnins.2022.1028929.
[5]
LECLER A, BOUZAD C, DESCHAMPS R, et al. Optimizing 3D FLAIR to detect MS lesions: pushing past factory settings for precise results[J]. J Neurol, 2019, 266(11): 2786-2795. DOI: 10.1007/s00415-019-09490-y
[6]
KIM H G, OH S W, HAN D, et al. Accelerated 3D T2-weighted images using compressed sensing for pediatric brain imaging[J]. Neuroradiology, 2022, 64(12): 2399-2407. DOI: 10.1007/s00234-022-03028-2.
[7]
LIN Z, GUO Z, QIU L, et al. The applied research of MRI with ASSET-EPI-FLAIR combined with 3D TOF MRA sequences in the assessment of patients with acute cerebral infarction[J]. Acta Radiol, 2016, 57: 1515-1523. DOI: 10.1177/0284185116628338.
[8]
IKEDA H, OHNO Y, MURAYAMA K, et al. Compressed sensing and parallel imaging accelerated T2 FSE sequence for head and neck MR imaging: Comparison of its utility in routine clinical practice[J/OL]. Eur J Radiol, 2020, 135: 109501 [2023-08-10]. https://pubmed.ncbi.nlm.nih.gov/33395594/. DOI: 10.1016/j.ejrad.2020.109501.
[9]
MUTHUPILLAI R, DOUGLAS E, HUBER S, et al. Direct comparison of sensitivity encoding (SENSE) accelerated and conventional 3D contrast enhanced magnetic resonance angiography (CE-MRA) of renal arteries: effect of increasing spatial resolution[J]. J Magn Reson Imaging, 2010, 31: 149-159. DOI: 10.1002/jmri.22002.
[10]
HAMILTON J, FRANSON D, SEIBERLICH N. Recent advances in parallel imaging for MRI[J]. Prog Nucl Mag Res Sp, 2017, 101: 71-95. DOI: 10.1016/j.pnmrs.2017.04.002.
[11]
李贝贝, 张浩楠, 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 Imaging Technol, 20222, 38(11): 1730-1734. DOI: 10.13929/j.issn.1003-3289.2022.11.031.
[12]
VIDYA SHANKAR R, HU H H, BIKKAMANE JAYADEV N, et al. 2-D magnetic resonance spectroscopic imaging of the pediatric brain using compressed sensing[J]. Pediatr Radiol, 2019, 49: 1798-1808. DOI: 10.1007/s00247-019-04495-1.
[13]
MUSSARD E, HILBERT T, FORMAN C, et al. Accelerated MP2RAGE imaging using Cartesian phyllotaxis readout and compressed sensing reconstruction[J]. Magnet Reson Med, 2020, 84: 1881-1894. DOI: 10.1002/mrm.28244.
[14]
VRANIC J E, CROSS N M, WANG Y, et al. Compressed sensing-sensitivity encoding (CS-SENSE) accelerated brain imaging: Reduced scan time without reduced image quality[J]. Am J Neuroradiol, 2019, 40(1): 92-98. DOI: 10.3174/ajnr.A5905.
[15]
TOLEDANO-MASSIAH S, SAYADI A, DE BOER R, et al. Accuracy of the compressed sensing accelerated 3D-FLAIR sequence for the detection of MS plaques at 3T[J]. Am J Neuroradiol, 2018, 39(3): 454-458. DOI: 10.3174/ajnr.A5517.
[16]
TIAN C, FEI L, ZHENG W, et al. Deep learning on image denoising: An overview[J]. Neural Networks, 2020, 131: 251-275. DOI: 10.1016/j.neunet.2020.07.025.
[17]
ANWAR S M, MAJID M, QAYYUM A, et al. Medical image analysis using convolutional neural networks: a review[J/OL]. J Med Syst, 2018, 42: 226 [2023-08-10]. https://pubmed.ncbi.nlm.nih.gov/30298337/. DOI: 10.1007/s10916-018-1088-1.
[18]
KIDOH M, SHINODA K, KITAJIMA M, et al. Deep learning based noise reduction for brain MR imaging: tests on phantoms and healthy volunteers[J]. Magn Reson Med Sci, 2020, 19: 195-206. DOI: 10.2463/mrms.mp.2019-0018.
[19]
SARTORETTI E, SARTORETTI-SCHEFER S, VAN SMOORENBURG L, et al. Contrast-enhanced T1-weighted head and neck MRI: Prospective intraindividual image quality comparison of spiral GRE, cartesian GRE, and cartesian TSE sequences[J]. Am J Roentgenol, 2022, 218: 132-139. DOI: 10.2214/AJR.21.26413.
[20]
YOU S H, KIM B, KIM B K. Transient global amnesia: Signal alteration in 2D/3D T2-FLAIR sequences[J]. Clin Imag, 2021, 78: 154-159, DOI: 10.1016/j.clinimag.2021.03.029.
[21]
WANG K Y, URIBE T A, LINCOLN C M. Comparing lesion detection of infratentorial multiple sclerosis lesions between T2-weighted spin-echo, 2D-FLAIR, and 3D-FLAIR sequences[J]. Clin Imag, 2018, 1: 229-234. DOI: 10.1016/j.clinimag.2018.05.017.
[22]
LIU T, WANG Y, XU Z, et al. 3D Cube FLAIR plus HyperSense compressed sensing is superior to 2D T2WI FLAIR scanning regarding image quality, resolutionspatial, detection rate for cortical microinfarcts[J/OL]. Medicine, 2022, 101(33): e28659 [2023-08-10]. https://pubmed.ncbi.nlm.nih.gov/35984121/. DOI: 10.1097/MD.0000000000028659.
[23]
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. DOI: 10.12015/issn.1674-8034.2022.04.018.
ZHANG Y Q, DING J L, GONG T, 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. DOI: 10.12015/issn.1674-8034.2022.04.018.
[24]
DEL GRANDE F, DELCOGLIANO M, GUGLIELMI R, et al. Fully automated 10-Minute 3D CAIPIRINHA SPACE TSE MRI of the knee in adults: A multicenter, multireader, multifield-strength validation study[J]. Invest Radiol, 2018, 53: 689-697. DOI: 10.1097/RLI.0000000000000493.
[25]
HOU B, LI Y, XIONG Y, et al. Comparison of CAIPIRINHA-accelerated 3D fat-saturated-SPACE MRI with 2D MRI sequences for the assessment of shoulder pathology[J]. Eur Radiol, 2021, 32: 593-601. DOI: 10.1007/s00330-021-08183-3.
[26]
SHENG R F, ZHENG L Y, JIN K P, et al. Single-breath-hold T2WI liver MRI with deep learning-based reconstruction: A clinical feasibility study in comparison to conventional multi-breath-hold T2WI liver MRI[J]. Magn Reson Imaging, 2021, 81: 75-81. DOI: 10.1016/j.mri.2021.06.014.
[27]
WANG S, CAO G, WANG Y, et al. Review and Prospect: Artificial Intelligence in Advanced Medical Imaging[J/OL]. Front Radiol, 2021, 1: 781868 [2023-08-10]. https://pubmed.ncbi.nlm.nih.gov/37492170/. DOI: 10.3389/fradi.2021.781868.
[28]
ZHANG Y, PENG W, XIAO Y, et al. Rapid 3D breath-hold MR cholangiopancreatography using deep learning-constrained compressed sensing reconstruction[J]. Eur Radiol, 2023, 33(4): 2500-2509. DOI: 10.1007/s00330-022-09227-y.
[29]
ZHAO Y, PENG C, WANG S, et al. The feasibility investigation of AI -assisted compressed sensing in kidney MR imaging: an ultra-fast T2WI imaging technology[J/OL]. BMC Med Imaging, 2022, 22(1): 119 [2023-08-10]. https://pubmed.ncbi.nlm.nih.gov/35787673/. DOI: 10.1186/s12880-022-00842-1.
[30]
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]. J Pain Res, 2023, 16: 257-267. DOI: 10.2147/JPR.S388219.
[31]
LIU K, XI B, SUN H, et al. The clinical feasibility of artificial intelligence-assisted compressed sensing single-shot fluid-attenuated inversion recovery (ACS-SS-FLAIR) for evaluation of uncooperative patients with brain diseases: comparison with the conventional T2-FLAIR with parallel imaging[J]. Acta Radiol, 2023, 64(5): 1943-1949. DOI: 10.1177/02841851221139125.
[32]
YANG F, PAN X, ZHU K, et al. Accelerated 3D high-resolution T2-weighted breast MRI with deep learning constrained compressed sensing, comparison with conventional T2-weighted sequence on 3.0 T[J/OL]. Eur J Radiol. 2022, 156: 110562 [2023-08-10]. https://pubmed.ncbi.nlm.nih.gov/36270194/. DOI: 10.1016/j.ejrad.2022.110562.
[33]
UEDA T, OHNO Y, YAMAMOTO K, et al. Compressed sensing and deep learning reconstruction for women's pelvic MRI denoising: Utility for improving image quality and examination time in routine clinical practice[J/OL]. Eur J Radiol, 2021, 134: 109430 [2023-08-10]. https://pubmed.ncbi.nlm.nih.gov/33276249/. DOI: 10.1016/j.ejrad.2020.109430.
[34]
潘珂, 刘倩倩, 唐玲玲, 等. 人工智能-压缩感知与压缩感知在膝关节MRI中的加速效率及图像质量评价研究[J]. 磁共振成像, 2022, 13(5): 94-98. DOI: 10.12015/issn.1674-8034.2022.05.017.
PAN K, LIU Q Q, TANG L L, et al. Study on acceleration efficiency and image quality of artificial intelligence compressed sensing and compressed sensing in knee MRI[J]. Chin J Magn Reson Imaging, 2022, 13(5): 94-98. DOI: 10.12015/issn.1674-8034.2022.05.017.
[35]
BUSTIN A, FUIN N, BOTNAR R M, et al. From compressed-sensing to artificial intelligence-based cardiac MRI reconstruction[J]. Front Cardiovasc Med, 2020: 7-17. DOI: 10.3389/fcvm.2020.00017.

上一篇 VMHC与ReHo在评价tDCS改善脑卒中后认知障碍中的应用价值
下一篇 心脏磁共振对比增强电影序列对合并微血管阻塞STEMI患者的诊断价值
  
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