分享:
分享到微信朋友圈
X
技术研究
基于人工智能压缩感知技术的鞍区多参数集成序列优化
罗贺丹 刘杨颖秋 张浩南 刘娜 张煜堃 袁畅 孙嘉忆 宋清伟 苗延巍

Cite this article as: LUO H D, LIU Y Y Q, ZHANG H N, et al. Optimization of multi-parameter MRI with flexible design sequences in the saddle area based on artificial intelligence-assisted compressed sensing[J]. Chin J Magn Reson Imaging, 2024, 15(12): 150-156.本文引用格式:罗贺丹, 刘杨颖秋, 张浩南, 等. 基于人工智能压缩感知技术的鞍区多参数集成序列优化[J]. 磁共振成像, 2024, 15(12): 150-156. DOI:10.12015/issn.1674-8034.2024.12.022.


[摘要] 目的 探究人工智能辅助下的压缩感知(artificial intelligence-assisted compressed sensing, ACS)技术对鞍区多参数集成序列(MULTI-parametric MR imaging with flexible design, MTP)成像的影响,并进行优化,筛选最合适的加速因子(acceleration factor, AF)。材料与方法 前瞻性纳入受试者41例。其中鞍区病变患者27例,健康志愿者14例。所有受试者使用3.0 T MRI行不同加速因子的MTP序列扫描,包括敏感性编码(sensitivity encoding, SENSE)技术的AF=3和ACS技术的AF=3、4、5和6(分别简写为SENSE3、ACS3、ACS4、ACS5和ACS6)。由MTP序列得到T1 map、R2* map、T2* map、T1WI、质子密度加权成像(proton density weighted imaging, PDWI)图像。两名观察者分别在不同AF序列的参数图及加权图上测量病灶、脑灰质信号强度(signal intensity, SI)及定量参数,测量脑白质SI、噪声强度(standard deviation, SD)及定量参数。分别计算两组不同AF之间的定量参数值,信噪比(signal-to-noise ratio, SNR)、对比噪声比(contrast-to-noise ratio, CNR)。根据图像伪影、病变显著性、灰白质分界清晰度采用五分评分法对图像质量进行主观评分。使用Kappa检验两观察者主观评分的一致性;采用组内相关系数(intra-class correlation coefficients, ICC)检验两名观察者客观测量数据的一致性。使用Friedman秩和检验或单因素ANOVA检验分析不同AF之间定量值、SNR、CNR及主观评分的差异。结果 两位观察者测量数据一致性良好(ICC:0.836~0.998,Kappa:0.839~0.909)。选择高年资观察者的主观评分及测量数据进行后续分析。两组各序列不同AF之间,定量值差异均无统计学意义(P>0.05);所测的SNR和CNR在不同的AF下差异具有统计学意义(P<0.05)。优化后不同AF(3~6)的ACS序列的扫描时间较SENSE3序列时间分别缩短21.21%、40.77%、52.62%、61.16%。与SENSE3序列相比,当ACS3、4时,T1WI图像和PDWI图像的SNR、CNR升高,差异具有统计学意义(P<0.05)。当ACS5、6时,PDWI图像的SNR升高,差异具有统计学意义(P<0.05);其余数据差异无统计学意义(P>0.05)。在不同AF之间,相比SENSE3序列图像,ACS5、6序列的灰白质分界清晰度主观评分降低,差异具有统计学意义;ACS6序列的病变显著性主观评分降低,差异具有统计学意义;其余数据差异无统计学意义(P>0.05)。结论 本研究结果显示采用ACS能够进一步优化MTP序列,综合时间和图像质量的平衡考虑,ACS4值得推荐,能获得质量可靠定量参数及定性图像。
[Abstract] Objective To investigate the impact of artificial intelligence-assisted compressed sensing (ACS) technology on the imaging of the MULTI-parametric MR imaging with flexible design sequences (MTP) in the saddle area, and to optimize and select the most suitable acceleration factor (AF).Materials and Methods Forty-one patients were prospectively included. There were 27 patients with sellar lesions and 14 healthy volunteers. All subjects underwent MTP sequential scanning with different AF using 3.0 T MRI. These included AF=3 for sensitivity encoding (SENSE) and AF=3, 4, 5 and 6 for ACS technology (abbreviated SENSE3, ACS3, ACS4, ACS5 and ACS6, respectively). The images of T1 map, R2* map, T2* map, T1WI and proton density weighted imaging (PDWI) were obtained from the MTP sequence. The signal intensity (SI) and quantitative parameters of the lesions and gray matter were measured by two observers on the parameter maps and weighted maps of different AF sequences, respectively. SI, standard deviation (SD) and quantitative parameters of white matter were measured. The quantitative parameter values, signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR), between the two groups of different AF were calculated respectively. According to image artifacts, lesion significance and gray matter demarcation clarity, five-point scoring method was used to evaluate the image quality. Kappa was used to test the consistency of subjective scores between the two observers. Intra-class correlation coefficients (ICC) were used to test the consistency of objective measurements between two observers. The differences in quantitative values, SNR, CNR and subjective scores among different AF were analyzed using Friedman rank sum test or single factor ANOVA test.Results There was a good agreement between the two observers (ICC: 0.836-0.998, Kappa: 0.839-0.909). The subjective scores and measurement data of senior observers were selected for follow-up analysis. There was no significant difference in quantitative values among different AF sequences (P>0.05). The measured SNR and CNR were different under different AF, and there was statistical significance (P<0.05). After optimization, the scanning time of ACS sequences with different AF (3-6) was reduced by 21.21%, 40.77%, 52.62% and 61.16%, respectively, compared with that of SENSE3 sequences. Compared with SENSE3 sequence, SNR and CNR of T1WI image and PDWI image were increased when ACS3 and 4 were used, and the difference was statistically significant (P<0.05). When ACS5, 6, the SNR of PDWI images increased, and the difference was statistically significant (P<0.05). There was no significant difference in other data (P>0.05). Compared with SENSE3 sequence images, the subjective score of gray-white matter demarcation clarity in ACS5 and 6 sequences was lower among different AFS, and the difference was statistically significant. The subjective score of lesion significance of ACS6 sequence decreased, and the difference was statistically significant. There was no significant difference in other data (P>0.05).Conclusions The results of this study show that using ACS can further optimize MTP sequences. Comprehensively considering time and image quality, ACS4 is recommended, with scan time reduced by 40.77%, and able to obtain reliable quantitative parameters and qualitative images.
[关键词] 多参数集成序列;压缩感知;人工智能;鞍区;磁共振成像
[Keywords] multi-parameter synthetic sequence;compressed sensing;artificial intelligence;saddle region;magnetic resonance imaging

罗贺丹 1, 2   刘杨颖秋 3   张浩南 1   刘娜 1   张煜堃 1   袁畅 1   孙嘉忆 1   宋清伟 1   苗延巍 1*  

1 大连医科大学附属第一医院放射科,大连116011

2 大连市公共卫生临床中心放射科,大连 116031

3 淄博市中心医院放射科,淄博 255000

通信作者:苗延巍,E-mail: ywmiao716@163.com

作者贡献声明:苗延巍设计本研究的方案,对稿件重要内容进行了修改;罗贺丹起草和撰写稿件,获取、分析本研究的数据;刘杨颖秋、张浩南、刘娜、张煜堃、孙嘉忆、袁畅、宋清伟对本研究的构思和设计有实质性贡献,获取、分析或解释本研究的数据,对稿件重要内容进行了修改;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:2024-06-25
接受日期:2024-12-10
中图分类号:R445.2  R743.3 
文献标识码:A
DOI: 10.12015/issn.1674-8034.2024.12.022
本文引用格式:罗贺丹, 刘杨颖秋, 张浩南, 等. 基于人工智能压缩感知技术的鞍区多参数集成序列优化[J]. 磁共振成像, 2024, 15(12): 150-156. DOI:10.12015/issn.1674-8034.2024.12.022.

0 引言

       多参数集成MRI序列(synthetic magnetic resonance imaging, SyMRI)一次扫描可重建出多组参数定量图像及定性参数图像,可以对目标部位进行多维度评估,同时获得T1值、T2*值和R2*值等量化参数[1, 2]。MTP(MULTI-parametric MR imaging with flexible design, MTP)就是其中一种[3]。和常规MRI序列相比,SyMRI已被证明在临床应用上具有明显优势,而且具有较好的准确性和可重复性[4, 5, 6],已经用于正常人群和脑肿瘤等重大疾病的定量评价[7, 8, 9]。然而,传统的SyMRI序列是二维序列,分辨率较低[10, 11],而三维同体素MTP序列虽然具有高分辨率、各向同性体素等诸多优点,但是其在未采用加速技术的情况下扫描时间过长却是值得关注的问题。

       压缩感知(compressed sensing, CS)技术采用稀疏变换从欠采样K空间中恢复图像信息,缩短了患者的检查时间,从而得到广泛的应用[12, 13, 14]。CS技术的图像信息可以通过稀疏变换从部分获取的数据中重建,由于有限的稀疏变换算法,很难从随机且高度欠采样的K空间数据中重建未损坏的图像或微小的解剖结构,稀疏性不足可能会导致类似于噪声的卷褶伪影,尤其是在加速因子(acceleration factor, AF)过高的情况下[15, 16]。人工智能辅助下的压缩感知(artificial intelligence-assisted compressed sensing, ACS)技术仍然采用CS技术,融合了压缩感知、并行采集和半傅里叶变换等多种重建算法,突破了常规单一加速方式所带来的限制,利用CS技术获取的欠采样K空间数据可以在深度学习神经网络的帮助下生成全采样数据。因此,与CS技术相比,ACS技术对原始K空间数据的最低采集需求可以进一步降低,从而产生更高的加速效率[17],可在保证图像质量的前提下缩短扫描时间[18, 19]。目前ACS技术已在颅脑、腹部、骨肌系统中应用并取得良好的效果[20, 21, 22]。然而,目前国内外尚未有基于ACS技术对SyMRI序列优化的报道。因此,本研究基于ACS对鞍区MTP序列进行优化,平衡扫描时间和图像质量,并监测定量参数的稳定性,缩短检查时间,选择最佳的AF。本研究为鞍区病变影像定性和定量诊断提供新方法,可提高检查效率,并提供给临床更多影像学信息。

1 材料与方法

1.1 研究对象

       前瞻性纳入于2022年8月至2024年7月在大连医科大学附属第一医院就诊并行MRI检查就诊的鞍区占位病变患者27例。同期社会招募健康志愿者14例。健康志愿者纳入标准:(1)既往无颅脑手术史;(2)无MRI检查禁忌证。排除标准:(1)因扫描序列时间长等因素而配合不佳,无法进行MRI检查;(2)MRI图像存在明显伪影无法诊断。鞍区病变组纳入标准:(1)鞍区占位性病变患者;(2)无MRI检查禁忌证。排除标准:(1)因总体扫描序列时间长等因素而配合不佳,无法进行MRI检查;(2)MR图像存在明显伪影无法诊断;(3)瘤体中含有明显囊变、坏死、出血、钙化成分。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,批准文号:PJ-KS-KY-2022-274。全体受试者均签署了知情同意书。

1.2 仪器与方法

       所有受试者均使用联影uMR OMEGA 3.0 T MR(中国)设备进行MRI检查,采用32通道相控阵头部线圈扫描。所有受试者均行轴位MTP序列扫描,分别行不同AF的MTP序列扫描,包括敏感性编码(sensitivity encoding, SENSE)的AF=3和ACS技术的AF=3、4、5和6(分别简写为SENSE3、ACS3、ACS4、ACS5和ACS6)。其他常规序列包括冠状位和矢状位的T2WI序列、冠状位和矢状位的T1WI序列以及注射对比剂后T1WI动态增强序列。具体扫描参数见表1

表1  各序列扫描参数
Tab. 1  Scan parameters of each sequence

1.3 数据测量及图像分析

1.3.1 MRI图像测量

       所有MTP原始数据传至联影工作站(uWS MR Version R005,中国)自动获得T1 map、R2* map、T2* map定量图像及T1WI、质子密度加权成像(proton density weighted imaging, PDWI)定性图像。由具有5年MRI操作经验的主管技师及9年MRI诊断经验的副主任医师在不同AF序列的T1 map、R2* map、T2* map、T1WI、PDWI图像上,避开囊变、坏死、出血、钙化成分,观察增强强化病灶,在所对应实质位置的目标图像上勾画3个圆形感兴趣区(region of interest, ROI),ROI大小为10~30 mm2,测量其信号强度(signal intensity, SI)及定量参数值,取平均值。在脑白质、灰质分别勾画3个ROI,大小为5~15 mm2,测量脑灰质的SI及定量参数,脑白质SI、定量参数及噪声强度(standard deviation, SD)(图1)。

       分别计算各图像的信噪比(signal-to-noise ratio, SNR)、对比噪声比(contrast-to-noise ratio, CNR),见公式(1)~(2)。

图1  男,46岁,病理诊断为垂体瘤,红圈分别为患者在MTP ACS4序列上T1 map图像勾画病灶、脑白质及脑灰质ROI的示意图。ACS:人工智能辅助下的压缩感知;ROI:感兴趣区。
Fig. 1  A 46-year-old male is pathologically diagnosed as a pituitary tumor. The red circles represent the patient's schematic drawings of lesions, white matter and gray matter ROI on the MTP ACS4 sequence using T1 map images. ACS: artificial intelligence-assisted compressed sensing; ROI: region of interest.

1.3.2 主观评分

       两位观察者根据伪影、灰白质分界清晰度,采用五分评分法分别对健康志愿者及鞍区病变患者(共41例)图像质量进行主观评分;根据病变显著性(27例)采用五分评分法对鞍区病变患者图像质量进行主观评分,评分标准如下[23, 24]

       伪影评分:伪影包括图像颗粒度、局部实质信号异常、搏动伪影、脑脊液信号强度不均匀以及运动伪影。1分,伪影严重,图像不可使用;2分,伪影明显,不能用于诊断;3分,存在明确伪影,但不影响诊断;4分,伪影极轻微,需仔细分辨方可显示;5分,无伪影。

       灰白质分界清晰度评分:以尾状核和豆状核分界以及皮层和皮层下白质分界的清晰度来评价,1分,无明确分界;2分,分界较差,不可分辨;3分,分界良好,可分辨;4分,分界较清晰;5分,分界清晰。

       鞍区病变显著性评分:1分,不能显示;2分,模糊,较难分辨;3分,模糊,可分辨;4分,较清晰;5分,非常清晰。

1.4 统计学分析

       采用G*Power 3.1.9.7(杜赛尔多夫大学,德国)软件进行样本量估算,采用单因素多水平(受试内)检验(F tests-ANOVA: Repeated measures, within factors),取显著性水平α=0.05,检验力1-β=0.8,效应量Effect size f=0.25,采取双侧检验,基于不同AF对图像质量的影响估算样本量。使用统计软件包SPSS 26.0软件对数据进行分析。使用Kappa检验评估两观察者主观评分的一致性;采用组内相关系数(intra-class correlation coefficients, ICC)检验两名观察者客观测量数据的一致性。Kappa或ICC值>0.70为一致性良好,Kappa或ICC值0.40~0.70为一致性一般;Kappa或ICC值<0.40为一致性差。采用Shapiro-Wilk检验各定量数据是否符合正态分布,符合正态分布进行单因素ANOVA分析检验;不符合正态分布的数据使用Friedman秩和检验;分析不同AF之间定量参数值、SNR、CNR及主观评分的差异,两种方法均对P值进行Bonferroni多重校正,校正后P<0.05为差异具有统计学意义。

图3  男,46岁,3A~3C分别为T1 map、R2* map、T2* map伪彩图。色条所示颜色由紫色到红色表示定量值由低到高。
Fig. 3  Male, 46 years old, 3A-3C are the pseudo-color maps of T1 map, R2* map, T2* map, respectively, where the color from purple to red indicates the quantitative value from low to high as shown in the color bar.

2 结果

2.1 样本量分析结果

       通过样本量估算得到需要样本量27例,本研究实际纳入病例共41例,符合样本量要求。

2.2 一般资料

       健康志愿者的年龄为(51.14±11.49)岁,男5例;女9例;鞍区病变患者的年龄为(49.59±11.74)岁,男10例,女17例。健康志愿者及鞍区病变患者性别及年龄差异均无统计学意义(P=0.689、0.763)。

2.3 一致性分析

       两名观察者图像质量主观评分及客观测量均具有良好的一致性(ICC:0.836~0.998,Kappa:0.839~0.909)。选择高年资观察者的主观评分及测量值进行后续分析。

2.4 各序列客观测量值不同AF间的差异

       两组各序列不同AF之间,定量值差异均无统计学意义(P>0.05);所测的SNR和CNR在不同的AF下,差异具有统计学意义(P<0.05,图23)。

       两两比较显示,与SENSE3序列相比,ACS3、ACS4序列在T1WI图像和PDWI图像上的SNR、CNR升高,差异具有统计学意义(P<0.05);ACS5、ACS6在PDWI图像的SNR升高,差异具有统计学意义(P<0.05)。其余数据差异无统计学意义(P>0.05,表23图4)。

图2  鞍区病变患者,女,60岁,2A~2E分别为SENSE3、ACS3、ACS4、ACS5和ACS6序列T1 map图像。SENSE:敏感度编码技术;ACS:人工智能辅助下的压缩感知。
Fig. 2  Patients with sellar lesions, female, 60 years old, 2A-2E are SENSE3, ACS3, ACS4, ACS5 and ACS6 sequence images, respectively. SENSE: sensitivity encoding; ACS: artificial intelligence-assisted compressed sensing.
图4  加权图像的SNR、CNR两两比较柱状图;4A:不同AF之间PDWI和T1WI图像的SNR两两比较柱状图;4B:不同AF之间PDWI和T1WI图像的CNR两两比较柱状图。*代表与SENSE3序列相比较差异具有统计学意义;P值经过Bonferroni校正。SNR:信噪比;CNR:对比噪声比;AF:加速因子;PDWI:质子密度加权成像;ACS:人工智能辅助下的压缩感知;SENSE:敏感度编码技术。
Fig. 4  Column chart of SNR and CNR pairwise comparison of weighted images. 4A: The bar chart of SNR pairwise comparison of PDWI and T1WI images between different AF; 4B: The histogram of CNR pairwise comparison between PDWI and T1WI images between different AF. * indicates statistically significant difference compared with SENSE3 sequence; The P value is corrected by Bonferroni. SNR: signal-to-noise ratio; CNR: contrast-to-noise ratio; AF: acceleration factor; PDWI: proton density weighted imaging; ACS: artificial intelligence-assisted compressed sensing; SENSE: sensitivity encoding.
表2  各序列不同AF之间定量值、客观测量的差异
Tab. 2  Differences in quantitative values and objective measurements among different AF sequences
表3  不同ACS序列和SENSE序列两两比较
Tab. 3  Pairwise comparison of different ACS sequences and SENSE sequences

2.5 各序列主观评分不同AF间的差异

       两组各序列不同AF之间,灰白质分界清晰度及病变显著性评分差异具有统计学意义(P<0.05)。对不同AF序列进行两两比较,与SENSE3序列相比,ACS5、6序列的灰白质分界清晰度主观评分降低,差异具有统计学意义;ACS6序列的病变显著性主观评分降低,差异具有统计学意义(P<0.05,表4图5)。

图5  主观评分两两比较柱状图;5A:不同AF之间伪影及灰白质分界主观评分两两比较柱状图;5B:不同AF之间病变显著性两两比较柱状图。*代表与SENSE3比较差异具有统计学意义;P值经过Bonferroni校正。SENSE:敏感度编码技术;ACS:人工智能辅助下的压缩感知;AF为加速因子。
Fig. 5  Bar chart of pairwise comparison of subjective scores; 5A: The bar chart of pairwise comparison of subjective scores of artifacts and gray-white matter demarcation between different AF; 5B: The bar graph of pairwise comparison of lesion significance between different AF. * indicates statistically significant difference compared with SENSE3 sequence; The P value is corrected by Bonferroni. SENSE: sensitivity encoding; ACS: artificial intelligence-assisted compressed sensing; AF: acceleration factor.
表4  各序列不同AF之间主观评分的差异
Tab. 4  Differences in subjective scores among different AF sequences

2.6 不同AF序列较SENSE序列缩短的时间

       鞍区病变组:不同AF扫描时间分别为363 s(SENSE3)、286 s(ACS3)、215 s(ACS4)、172 s(ACS5)、141 s(ACS6)。优化后不同AF(3~6)的ACS序列较SENSE3序列时间分别缩短21.21%、40.77%、52.62%、61.16%。

3 讨论

       本研究对鞍区病变患者不同AF状态下的ACS MTP序列进行优化,在不影响定量值测量和图像质量的情况下,选择最适合临床应用的AF,结果表明当ACS AF=4时,可在保证图像的前提下缩短扫描时间40.77%。目前国内外尚没有对于MTP序列在颅脑部鞍区进行优化的研究,本研究证明优化后的MTP序列可提高检查效率,可对其颅脑部鞍区病变进行定性及定量评估。

3.1 MTP序列的优势及应用价值

       本研究所用的MTP序列基于梯度回波(gradient echo, GRE)序列,它由两个TR,双翻转角和6个不同的TE组成。GRE采集使用信号扰相技术来收集非相干稳态信号,来进行三维高分辨率(体素尺寸<1 mm)成像,可以与加速技术兼容,该序列校正了B1场不均匀性,减少颅底磁敏感伪影,从而提高了准确性和可重复性[3]。该序列采集后不只生成定量图像也可以生成T1WI、PDWI图像,从而避免了常规临床环境中需要额外扫描常规加权成像,具有一定优势。既往研究在健康志愿者和病变患者年龄分层的研究中证明了该序列的可行性[25, 26]。同时,MTP在肿瘤的分级诊断及生物标记物的预测方面取得了较好的成果[27, 28, 29],可对肿瘤及病变进行定量评估研究,既往研究[30]表明,垂体瘤患者在术前进行多参数集成序列核磁成像,促性腺激素垂体瘤的T1和T2值显著高于非促性腺激素垂体瘤,并且T1值和T2值联合诊断可以区分这两种亚型,非促性腺激素垂体瘤更有可能侵犯海绵窦,因此,尽早评估这两种亚型对该疾病预后具有一定价值。该序列生成的PDWI图像可以观察病变对海绵窦内壁的浸润,优于检测海绵窦浸润的传统诊断标准,为临床提供准确的术前图像[31]。但该序列未优化前扫描时间长达6 min,增加了患者不适和运动伪影增加,因此对该序列进行优化非常必要。

3.2 ACS技术应用于MTP序列的可行性及临床价值

       ACS技术引入了基于深度学习神经网络的人工智能(artificial intelligence, AI)模块,以加速K空间的填充,是一种集成深度神经网络、半傅立叶成像、并行成像和压缩感知的新技术,可以抑制噪声、减少伪影,在不牺牲图像质量的情况下加速MRI数据采集,有效节省成像时间并实现成像质量之间的平衡[32, 33]。成像的可靠性相比单独基于AI的加速方法也得到了提升。可以纠正图像产生的错误,从而抑制噪声、减少伪影,可以实现更高的加速水平。应用ACS技术图像的图像质量非常适合临床诊断,SNR和CNR都有所提高。本研究纳入主观和客观评价因素,对图像同时从三个不同的方面对图像进行主观评估;客观评估不仅包含对定性图像的评估,我们对不同AF序列之间的定量值也进行了差异性评估,并将这些序列应用于临床,使得评价更加全面,结果更可靠。在客观测量中,随着AF的增加,定量值之间的差异无统计学意义,但图像质量逐渐降低,而扫描时间缩短,这与既往研究相同[34, 35, 36]。ACS3和ACS4均具有作为MTP序列优化序列的潜力。与SENSE3序列相比,ACS3序列可以生成与SENSE3序列相同甚至更好的图像质量,但缩短扫描时间有限,在客观测量中ACS4序列的SNR、CNR略低于ACS3,但可在节约时间和良好图像质量之间取得平衡。主观评估结果表明,ACS3至ACS6序列的图像伪影评分与SENSE3相当;其SNR、CNR随着AF的增加而降低。当AF增加到5或6时,虽然ACS5与ACS6序列的SNR、CNR未显著低于原始SENSE3序列,但在主观评价中,ACS5序列灰白质分界清晰度较SENSE3序列降低,差距具有统计学意义;ACS6序列灰白质分界清晰度及病变显著性评分较SENSE3序列均降低,差距具有统计学意义。因此,不作为优化的序列选择。当AF=4时,与SENSE3序列相比,TIWI、PDWI图像的SNR及CNR均升高,差异具有统计学意义;主观评分差异无统计学意义。将优化后的MTP序列应用于临床,与常规序列相比扫描时间可以缩短40.77%。使用ACS技术缩短了扫描时间,可得到与传统序列SENSE3相似甚至更佳的图像质量。可以有效节省成像时间并实现成像质量之间的平衡。因此,推荐使用ACS4序列作为优化序列进行MTP序列临床扫描。

3.3 本研究的局限性

       本研究存在一定的局限性。首先,本研究样本量较小,有待进一步扩大样本量;其次,本研究仅是单中心研究,其结果还需要经过多中心进一步验证;最后,本研究并未对鞍区多种病变进行分类分组进行定量测量研究,后续应扩大疾病种类将其定量值与临床、病理进行对照研究。

4 结论

       本研究利用ACS技术保持较高的图像质量和准确的定量分析的同时,可以减少MTP序列扫描时间,综合时间和成像质量的考虑,ACS4适用于颅脑部鞍区病变MTP成像,推荐将ACS4应用于临床。

[1]
GONÇALVES F G, SERAI S D, ZUCCOLI G. Synthetic brain MRI: review of current concepts and future directions[J]. Top Magn Reson Imaging, 2018, 27(6): 387-393. DOI: 10.1097/RMR.0000000000000189.
[2]
JI S, YANG D J, LEE J, et al. Synthetic MRI: technologies and applications in neuroradiology[J]. J Magn Reson Imaging, 2022, 55(4): 1013-1025. DOI: 10.1002/jmri.27440.
[3]
YE Y Q, LYU J Y, HU Y C, et al. MULTI-parametric MR imaging with fLEXible design (MULTIPLEX)[J]. Magn Reson Med, 2022, 87(2): 658-673. DOI: 10.1002/mrm.28999.
[4]
刘雅文, 牛海军, 尹红霞, 等. 合成MRI与传统定量方法对T1、T2弛豫值测定的模体验证对比研究[J]. 磁共振成像, 2022, 13(4): 89-93. DOI: 10.12015/issn.1674-8034.2022.04.016.
LIU Y W, NIU H J, YIN H X, et al. A comparative study on phantom verification of T1 and T2 relaxation values determined by synthetic MRI and conventional mapping methods[J]. Chin J Magn Reson Imag, 2022, 13(4): 89-93. DOI: 10.12015/issn.1674-8034.2022.04.016.
[5]
刘辉明, 尹国平, 别非, 等. 对比集合序列与常规序列头部MR图像质量[J]. 中国医学影像技术, 2019, 35(2): 268-271. DOI: 10.13929/j.1003-3289.201805163.
LIU H M, YIN G P, BIE F, et al. Comparison of image quality of brain between conventional MR sequence and compilation sequence[J]. Chin J Med Imag Technol, 2019, 35(2): 268-271. DOI: 10.13929/j.1003-3289.201805163.
[6]
HAGIWARA A, HORI M, COHEN-ADAD J, et al. Linearity, bias, intrascanner repeatability, and interscanner reproducibility of quantitative multidynamic multiecho sequence for rapid simultaneous relaxometry at 3 T: a validation study with a standardized phantom and healthy controls[J]. Invest Radiol, 2019, 54(1): 39-47. DOI: 10.1097/RLI.0000000000000510.
[7]
ANDICA C, HAGIWARA A, HORI M, et al. Automated brain tissue and myelin volumetry based on quantitative MR imaging with various in-plane resolutions[J]. J De Neuroradiol, 2018, 45(3): 164-168. DOI: 10.1016/j.neurad.2017.10.002.
[8]
LEE S M, CHO H H, YOU S K, et al. Time-saving synthetic magnetic resonance imaging protocols for pediatric neuroimaging: impact of echo train length and bandwidth on image quality[J]. Pediatr Radiol, 2022, 52(12): 2401-2412. DOI: 10.1007/s00247-022-05389-5.
[9]
王佳男, 朱婧怡, 张博, 等. 脑良恶性胶质瘤实质与瘤周T1、T2值差别的研究[J]. 磁共振成像, 2020, 11(6): 416-421. DOI: 10.12015/issn.1674-8034.2020.06.004.
WANG J N, ZHU J Y, ZHANG B, et al. The study of T1 and T2 values in differentiating of benign and malignant gliomas[J]. Chin J Magn Reson Imag, 2020, 11(6): 416-421. DOI: 10.12015/issn.1674-8034.2020.06.004.
[10]
FUJITA S, HAGIWARA A, HORI M, et al. 3D quantitative synthetic MRI-derived cortical thickness and subcortical brain volumes: scan-rescan repeatability and comparison with conventional T1-weighted images[J]. J Magn Reson Imaging, 2019, 50(6): 1834-1842. DOI: 10.1002/jmri.26744.
[11]
FUJITA S, HAGIWARA A, OTSUKA Y, et al. Deep learning approach for generating MRA images from 3D quantitative synthetic MRI without additional scans[J]. Invest Radiol, 2020, 55(4): 249-256. DOI: 10.1097/RLI.0000000000000628.
[12]
LEE H K, SONG J S, JANG W, et al. Improved single breath-hold SSFSE sequence for liver MRI based on compressed sensing: evaluation of image quality compared with conventional T2-weighted sequences[J/OL]. Diagnostics, 2022, 12(9): 2164 [2024-06-24]. https://pubmed.ncbi.nlm.nih.gov/36140565/. DOI: 10.3390/diagnostics12092164.
[13]
BISCHOFF L M, KATEMANN C, ISAAK A, et al. T2 turbo spin echo with compressed sensing and propeller acquisition (sampling k-space by utilizing rotating blades) for fast and motion robust prostate MRI: comparison with conventional acquisition[J]. Invest Radiol, 2023, 58(3): 209-215. DOI: 10.1097/RLI.0000000000000923.
[14]
QIU J X, LIU J, BI Z X, et al. An investigation of 2D spine magnetic resonance imaging (MRI) with compressed sensing (CS)[J]. Skeletal Radiol, 2022, 51(6): 1273-1283. DOI: 10.1007/s00256-021-03954-x.
[15]
YOON S, PARK S H, HAN D. Uncover this tech term: compressed sensing magnetic resonance imaging[J]. Korean J Radiol, 2023, 24(12): 1293-1302. DOI: 10.3348/kjr.2023.0743.
[16]
BUSTIN A, FUIN N, BOTNAR R M, et al. From compressed-sensing to artificial intelligence-based cardiac MRI reconstruction[J/OL]. Front Cardiovasc Med, 2020, 7: 17 [2024-06-24]. https://pubmed.ncbi.nlm.nih.gov/32158767/. DOI: 10.3389/fcvm.2020.00017.
[17]
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 [2024-06-24]. https://pubmed.ncbi.nlm.nih.gov/36744117/. DOI: 10.2147/JPR.S388219.
[18]
KOZAK B M, JAIMES C, KIRSCH J, et al. MRI techniques to decrease imaging times in children[J]. Radiographics, 2020, 40(2): 485-502. DOI: 10.1148/rg.2020190112.
[19]
LIU H B, DENG D L, ZENG W L, et al. AI-assisted compressed sensing and parallel imaging sequences for MRI of patients with nasopharyngeal carcinoma: comparison of their capabilities in terms of examination time and image quality[J]. Eur Radiol, 2023, 33(11): 7686-7696. DOI: 10.1007/s00330-023-09742-6.
[20]
LIU K, XI B, SUN H T, 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.
[21]
LI H L, HU C L, YANG Y, et al. Single-breath-hold T2WI MRI with artificial intelligence-assisted technique in liver imaging: As compared with conventional respiratory-triggered T2WI[J/OL]. Magn Reson Imaging, 2022, 93: 175-180 [2024-06-24]. https://pubmed.ncbi.nlm.nih.gov/35987419/. DOI: 10.1016/j.mri.2022.08.012.
[22]
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 [2024-06-24]. https://pubmed.ncbi.nlm.nih.gov/37857868/. DOI: 10.1186/s41747-023-00374-5.
[23]
郑作锋, 张东坡, 毛磊, 等. 颅脑集成MRI与常规MRI图像质量的对比研究[J]. 磁共振成像, 2019, 10(5): 361-365. DOI: 10.12015/issn.1674-8034.2019.05.009.
ZHENG Z F, ZHANG D P, MAO L, et al. Comparative study of image quality between synthetic MRI and conventional MRI[J]. Chin J Magn Reson Imag, 2019, 10(5): 361-365. DOI: 10.12015/issn.1674-8034.2019.05.009.
[24]
吕青青, 廖俊杰, 任娜娜, 等. 儿童颅脑集成MRI的图像质量研究[J]. 临床放射学杂志, 2023, 42(11): 1795-1799. DOI: 10.13437/j.cnki.jcr.2023.11.013.
LÜ Q Q, LIAO J J, REN N N, et al. The study of image quality at synthetic brian MRI in children[J]. J Clin Radiol, 2023, 42(11): 1795-1799. DOI: 10.13437/j.cnki.jcr2023.11.013.
[25]
VANDERHASSELT T, NAEYAERT M, WATTÉ N, et al. Synthetic MRI of preterm infants at term-equivalent age: evaluation of diagnostic image quality and automated brain volume segmentation[J]. AJNR Am J Neuroradiol, 2020, 41(5): 882-888. DOI: 10.3174/ajnr.A6533.
[26]
ZOU M S, ZHOU Q, LI R C, et al. Image quality using synthetic brain MRI: an age-stratified study[J]. Acta Radiol, 2023, 64(5): 2010-2023. DOI: 10.1177/02841851231152098.
[27]
KONAR A S, SHAH A D, PAUDYAL R, et al. Quantitative synthetic magnetic resonance imaging for brain metastases: a feasibility study[J/OL]. Cancers, 2022, 14(11): 2651 [2024-06-24]. https://pubmed.ncbi.nlm.nih.gov/35681631/. DOI: 10.3390/cancers14112651.
[28]
SACCO S, BALLATI F, GAETANI C, et al. Multi-parametric qualitative and quantitative MRI assessment as predictor of histological grading in previously treated meningiomas[J]. Neuroradiology, 2020, 62(11): 1441-1449. DOI: 10.1007/s00234-020-02476-y.
[29]
MAURER G D, TICHY J, HARTER P N, et al. Matching quantitative MRI parameters with histological features of treatment-Naïve IDH wild-type glioma[J/OL]. Cancers, 2021, 13(16): 4060 [2024-06-24]. https://pubmed.ncbi.nlm.nih.gov/34439213/. DOI: 10.3390/cancers13164060.
[30]
BAI Y, SHEN Y, CHEN R S, et al. Magnetic resonance fingerprinting for preoperative differentiation between gonadotroph and non-gonadotroph pituitary macroadenomas[J]. Eur Radiol, 2021, 31(11): 8420-8428. DOI: 10.1007/s00330-021-07950-6.
[31]
NAYAK A M, CHOUDHARI A, PATKAR D P, et al. Pituitary volume and iron overload evaluation by 3T MRI in thalassemia[J]. Indian J Pediatr, 2021, 88(7): 656-662. DOI: 10.1007/s12098-020-03629-w.
[32]
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.
[33]
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]. Eur Radiol, 2024, 34(5): 3046-3058. DOI: 10.1007/s00330-023-10368-x.
[34]
罗贺丹, 张浩南, 高冰冰, 等. 联合压缩感知技术优化脑神经一站式三维可调制翻转角超长回波链回聚脉冲(3D-MATRIX)成像[J]. 中国医学影像技术, 2023, 39(6): 905-909. DOI: 10.13929/j.issn.1003-3289.2023.06.024.
LUO H D, ZHANG H N, GAO B B, et al. United compressed sensing technology for optimizing one-stop threedimensional modulated flip angles technique in refocusing imaging with extended echo train (3D-MATRIX) imaging of cranial nerves[J]. Chin J Med Imag Technol, 2023, 39(6): 905-909. DOI: 10.13929/j.issn.1003-3289.2023.06.024.
[35]
张煜堃, 常佩佩, 高冰冰, 等. 3.0T 3D高分辨率MR血管壁成像显示脑小血管病豆纹动脉改变[J]. 中国医学影像技术, 2023, 39(5): 665-669. DOI: 10.13929/j.issn.1003-3289.2023.05.005.
ZHANG Y K, CHANG P P, GAO B B, et al. 3.0T three-dimensional high resolution vessel wall MRI for displaying lenticulostriate artery changes of cerebral small vessel disease[J]. Chin J Med Imag Technol, 2023, 39(5): 665-669. DOI: 10.13929/j.issn.1003-3289.2023.05.005.
[36]
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.
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 Imag, 2024, 15(4): 139-144, 152. DOI: 10.12015/issn.1674-8034.2024.04.022.

上一篇 IDEAL-IQ序列定量评估2型糖尿病大鼠腹腔实质脏器脂肪沉积及铁过载的应用研究
下一篇 胎盘梗死并胎儿颅脑急性缺血性改变MRI一例
  
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