分享:
分享到微信朋友圈
X
特别关注
基于深度学习重建和传统TSE序列在直肠癌磁共振检查的对比研究
胡思洁 范文文 滕泽 刘侃 童晓婉 江岳娈 刘鹏 郎宇 张红梅

Cite this article as: HU S J, FAN W W, TENG Z, et al. Study on the value of deep reconstruction technique in improving the image quality of magnetic resonance rectal cancer[J]. Chin J Magn Reson Imaging, 2024, 15(10): 30-35.本文引用格式:胡思洁, 范文文, 滕泽, 等. 基于深度学习重建和传统TSE序列在直肠癌磁共振检查的对比研究[J]. 磁共振成像, 2024, 15(10): 30-35. DOI:10.12015/issn.1674-8034.2024.10.006.


[摘要] 目的 评估深度学习重建(deep learning reconstruction, DL Recon)技术在提高直肠MRI快速自旋回波序列(turbo spin-echo, TSE)图像质量的价值。材料与方法 前瞻性纳入2023年9月至2024年1月中国医学科学院肿瘤医院病理诊断为直肠癌的初诊患者60例,对每位患者进行临床传统TSE序列及应用DL Recon技术的TSE序列(deep learning reconstruction-TSE, DL-TSE)扫描,并记录扫描时间,由两名影像科医师分别对两组图像(传统TSE、DL-TSE)的图像质量进行主观评价,采用“五分法”分别对病灶轮廓清晰度,图像伪影,病灶结构清晰度以及诊断信心进行评分;由两名影像科技师进行客观评价,分别计算DL-TSE和传统TSE图像的病灶信噪比(signal-to-noise ratio, SNR)以及病灶区域与周边组织的对比噪声比(contrast-to-noise ratio, CNR),采用配对样本t检验或配对样本非参数检验(Wilcoxon符号轶检验)进行统计学分析。结果 图像主观评价显示DL-TSE序列的病灶轮廓清晰度、图像伪影、病灶结构清晰度以及诊断信心的主观评分得分均优于传统TSE序列,且差异具有统计学意义(P<0.001)。图像客观评价显示DL-TSE序列和传统TSE序列的SNR分别为24.26(15.95, 42.79)、11.84(7.63, 18.88),差异有统计学意义(Z=-14.276,P<0.001);DL-TSE序列和传统TSE序列CNR分别为10.75(7.19, 15.63),5.47(3.72, 8.86),且差异有统计学意义(Z=-14.271,P<0.001)。DL-TSE序列的SNR及CNR相较于传统TSE序列均有明显提升。结论 DL-TSE序列通过采用原始K空间数据DL Recon算法,在保证图像质量和病变可检测性的情况下,可提升直肠癌患者序列图像SNR及CNR,并且可缩短36.6%扫描时间。
[Abstract] Objective To evaluate the value of deep learning reconstruction (DL Recon) technique in improving the image quality of rectal MRI turbo spin-echo (TSE) sequences.Materials and Methods Sixty new cases of rectal cancer diagnosed by pathology in the Chinese Academy of Medical Sciences from September 2023 to January 2024 were studied retrospectively. Each patient was subjected to a conventional TSE sequence and DL-TSE sequence, and the scanning time was recorded. Two imaging doctors had subjective evaluation for the two groups (conventional TSE, DL-TSE). The "five-point method" was used to score lesion contour clarity, the image artifacts, the clarity of the lesion and the reliability of the diagnosis, and the statistical description of the results was performed using the quartile interval M (Q25, Q75). The signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) between the DL-TSE and the conventional TSE images were computed by two imaging technicians. Paired sample t test was used for statistical analysis of the data conforming to normal distribution, and paired sample non-parametric test (Wilcoxon symbolic significance test) was used for statistical analysis of the data not conforming to normal distribution, and the results were statistically described by the quartile interval M (Q25, Q75).Results Sixty cases of rectal carcinoma aged 35-69 (53±10) years old were enrolled. The subjective evaluation results of conventional TSE sequences and DL-TSE sequences: The focal contour clarity, image artifacts, focal structure clarity and subjective score of diagnostic confidence of DL-TSE sequence were better than those of traditional TSE sequence, and the differences were statistically significant (P<0.001). Objective evaluation results of traditional TSE sequence and DL-TSE sequence images: The SNR of DL-TSE and conventional TSE sequences were 24.26 (15.95, 42.79) and 11.84 (7.63, 18.88). The CNR of DL-TSE and conventional TSE sequences were 10.75 (7.19, 15.63) and 5.47 (3.72, 8.86), the difference was statistically significant (Z=-14.271, P<0.001). The SNR and the CNR of the DL-TSE were obviously higher than those of conventional TSE sequences.Conclusions DL-TSE sequence uses the original K-space data DL Recon reconstruction algorithm to improve the SNR and CNR of sequence images of rectal cancer patients, and can shorten the scanning time by 36.6%, while ensuring the image quality and lesion detectability.
[关键词] 直肠癌;信噪比;对比噪声比;深度学习重建;磁共振成像
[Keywords] rectal cancer;signal-to-noise ratio;contrast-to-noise ratio;deep learning reconstruction;magnetic resonance imaging

胡思洁 1   范文文 1   滕泽 1   刘侃 1   童晓婉 1   江岳娈 2   刘鹏 1   郎宇 1   3   张红梅 1*  

1 国家癌症中心/国家肿瘤临床医学研究中心/中国医学科学院北京协和医学院肿瘤医院影像诊断科,北京 100021

2 西门子医疗系统有限公司北京分公司,北京 100102

3 西门子医疗系统有限公司,埃尔兰根,德国

通信作者:张红梅,E-mail: 13581968865@163.com

作者贡献声明:张红梅设计本研究的方案,对稿件重要内容进行修改;胡思洁起草和撰写稿件,获取、分析或解释本研究的数据;范文文、滕泽、刘侃、童晓婉、刘鹏、郎宇、江岳娈、Nickel MarcelDominik获取、分析和解释本研究的数据,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。


收稿日期:2024-03-29
接受日期:2024-08-02
中图分类号:R445.2  R735.37 
文献标识码:A
DOI: 10.12015/issn.1674-8034.2024.10.006
本文引用格式:胡思洁, 范文文, 滕泽, 等. 基于深度学习重建和传统TSE序列在直肠癌磁共振检查的对比研究[J]. 磁共振成像, 2024, 15(10): 30-35. DOI:10.12015/issn.1674-8034.2024.10.006.

0 引言

       直肠癌是常见的消化道恶性肿瘤之一,近年来,直肠癌在我国的发病率显著上升,从1972年的2.75增加到2019年的19.39(每10万人),因此直肠癌的筛查与治疗显得尤为重要[1, 2, 3, 4]。直肠MRI具有较高的软组织分辨率,在病变定位、解剖关系、准确的术前临床分期以及评估手术效果中起着至关重要的作用[5, 6],可以促进有效治疗策略的制订[7, 8, 9]

       在直肠MRI检查中多方位多对比度的成像是必不可少的,包括T1WI、T2WI、T2WI抑脂(fat saturated, FS)以及弥散加权成像(diffusion weighted imaging, DWI),其中快速自旋回波序列(turbo spin-echo, TSE)序列因其高病灶信噪比(signal to noise ratio, SNR)和高空间分辨率而成为标准序列[10, 11, 12]。然而,直肠MRI的TSE序列因其较长的采集时间容易受到肠蠕动引起的运动伪影和肠气-组织界面的磁感伪影影响。已有几种方法被用于缓解这些伪影,例如对前体壁皮下脂肪应用饱和带[13]、传统的平面内平行成像(parallel imaging, PI)方法[14]以及抗胆碱能药物的使用[15, 16]。然而,在实践中PI虽然减少了相位编码数以缩短采集时间,但是会导致SNR的降低,并可能使噪音增强或产生伪影。深度学习重建(deep learning reconstruction, DL Recon)技术出现之前,用于减少MRI扫描时间的技术有压缩传感(compressive sensing, CS)和并行采集成像等,尽管这些方法取得了一定的效果,但在高加速因子下仍面临噪声和伪影问题,可能会导致重建速度慢和重建图像质量不高,这些问题限制了这些技术在临床应用中的进一步推广和使用[17]

       DL Recon技术是一种利用深度学习(deep learning, DL)技术从欠采样的K空间数据中重建高质量MRI图像的方法,有望在提高SNR及对比噪声比(contrast-to-noise ratio, CNR)的同时实现更高的加速因子[18]。因此,我们假设DL-TSE序列可以生成与临床传统TSE序列图像相似的图像质量。此外,在更高的加速倍数和更低的平均次数下,TSE图像采集时间也可以显著减少。因此,本研究旨在评估使用了DL Recon缩短图像采集时间并提高图像质量的TSE序列应用于临床直肠MRI检查的可行性。

1 材料与方法

1.1 研究对象

       本研究为前瞻性研究,遵守《赫尔辛基宣言》,经中国医学科学院肿瘤医院伦理委员会批准(批准文号:NCC4435),全体受检者检查前均已签署知情同意书。纳入2023年9月至2024年1月中国医学科学院肿瘤医院就诊临床怀疑为直肠癌患者64例。纳入标准:(1)经临床医生诊断及相关检查怀疑为直肠癌的患者;(2)无MRI扫描禁忌证;(3)年满18周岁。排除标准:(1)病灶经手术或活检证实不是直肠癌;(2)病灶太小,无法在MRI影像中观察病变(图1)。

图1  直肠癌患者纳排流程图。
Fig. 1  Flow chart of patients with rectal cancer.

1.2 设备与参数

       所有60例患者均在西门子3.0 T磁共振扫描仪(MAGNETOM Prisma,西门子医疗,埃尔兰根,德国)上行常规直肠MRI扫描,包括矢状位T2WI(SAG T2WI)、冠状位T2WI(COR T2WI)、轴位T2WI FS(TRA FS T2WI)、轴位小视野T2WI(TRA SMALL T2WI)、轴位T1WI(TRA T1WI)的TSE序列,以及相应的DL-TSE序列。详细成像参数见表1

       患者在检查床中央仰卧位,双臂举过头顶。为保证图像质量,在扫描前将患者身上的所有金属首饰或衣物取下,进行标准化呼吸训练以减少呼吸运动伪影,并在无禁忌证的情况下向患者注射抗胆碱能药物(盐酸消旋山莨菪碱注射液,杭州民生药业股份有限公司,国药准字H33021707,规格1 mL∶10 mg,成人每次肌注10 mg)以减少肠道运动伪影。

表1  常规TSE序列与DL-TSE序列详细参数
Tab. 1  Detailed parameters of conventional TSE sequence and DL-TSE sequence

1.3 客观图像分析

       两名分别具有4年和6年影像技术经验的初级职称影像科技师(技师一和技师二)独立地在所有患者传统TSE序列中矢状T2WI、冠状T2WI、轴位小视野T2WI、轴位抑脂T2WI、轴位T1WI及DL-TSE序列中矢状T2WI、冠状T2WI、轴位小视野T2WI、轴位抑脂T2WI、轴位T1WI图像的直肠和邻近组织上放置了尽可能大的感兴趣区域(region of interest, ROI),并避开伪影区域。SNR和CNR的计算公式如下:SNR=SI/SD,CNR=|SIA-SIB|/SDnoise(A为直肠组织,B为直肠邻近组织),其中SI为信号强度,SD为信号标准差,SDnoise代表相应层面图像背景噪声信号标准差。

1.4 主观图像分析

       由两名分别具有8年和20年诊断经验的初级职称(医师一)和高级职称(医师二)的影像科医师采用盲读法分别对所有图像的整体图像质量进行主观评分,定量评估在Syngo MR工作站(Siemens Healthineers,埃尔兰根,德国)进行。两名医师独立评估了两个序列的病灶轮廓清晰度、图像伪影、病灶结构清晰度、诊断信心方面的图像质量(图2)。所有图像都处于盲读状态,使用Rickett量表5分标准:5为最佳,1为最差(表2)。

图2  直肠癌MRI常规TSE与DL-TSE图像主观评分示意图。2A、2C、2E、2G:常规TSE图像;2B、2D、2F、2H:DL-TSE图像。2A~2B:经病理诊断为直肠中分化腺癌的55岁男性患者SAG T2WI图像,直肠下端前壁可见长约2.7 cm异常信号,与传统TSE序列图像相比,DL-TSE序列病灶边缘更加清晰;2C~2D:经病理诊断为直肠中分化腺癌的55岁男性患者TRA SMALL T2WI图像,患者直肠中下段肠壁增厚,最厚处约0.9 cm,传统TSE序列图像信号不均匀,组织边缘不清,DL-TSE序列图像信号均匀,图像锐利;2E~2F:经病理诊断为直肠中分化腺癌的63岁女性患者TRA SMALL T2WI图像,直肠中段局部增厚,最厚处约0.8 cm,传统TSE序列病灶内部信号差,结构模糊,DL-TSE序列图像锐利,病灶结构清晰,获得了更好的整体图像质量;2G~2H:经病理诊断为直肠中分化腺癌的42岁女性患者TRA SMALL T2WI图像,肿瘤侵犯肠周的33%,最大厚度约1.0 cm,传统TSE序列图像中直肠组织与邻近组织边界不清,病灶轮廓模糊,DL-TSE序列图像病灶降低了噪声,增加了病灶轮廓的清晰度。TSE:快速自旋回波;DL-TSE:应用深度学习重建技术的TSE 序列;TRA SMALL:轴位小视野;SAG:矢状位。
Fig. 2  Schematic diagram of the subjective score of conventional MRI TSE and DL-TSE images for rectal cancer. 2A, 2C, 2E, 2G: conventional TSE images, 2B, 2D, 2F, 2H: DL-TSE images. 2A-2B: SAG T2WI images of a 55-year-old male patient pathologically diagnosed with moderately differentiated adenocarcinoma of the rectum. Abnormal signals of about 2.7 cm in length can be seen on the anterior wall of the lower rectum. Compared with traditional TSE sequence images, the lesion edges of DL-TSE sequence are clearer. 2C-2D: TRA SMALL T2WI images of a 55-year-old male patient with a pathological diagnosis of moderately differentiated adenocarcarcinoma of the rectum. The intestinal wall of the middle and lower rectum of the patient is thickened, with the thickest part being about 0.9 cm. The traditional TSE sequence image signals are uneven, and the tissue edges are unclear, while the DL-TSE sequence image signals are uniform and the images are sharp. 2E-2F: TRA SMALL T2WI images of a 63-year-old woman with pathologically diagnosed moderately differentiated adenocarcinoma of the rectum with local thickening in the middle section of the rectum. The thickest part is about 0.8 cm. Traditional TSE sequence lesions have poor internal signal and fuzzy structure, while DL-TSE sequence images are sharp and clear structure of lesions, and better overall image quality is obtained. 2G-2H: TRA SMALL T2WI images of a 42-year-old female patient with a pathological diagnosis of rectal medium differentiated adenocarcinoma. The tumor invaded 33% of the periintestinal tract, with a maximum thickness of about 1.0 cm. In traditional TSE sequence images, the boundary between rectal tissue and adjacent tissues is unclear, and the lesion contour is blurred, while the lesion in DL-TSE sequence images reduced noise. The definition of lesion contour is increased. TSE: turbo spin-echo; DL-TSE: deep learning reconstruction-TSE; TRA: transvers; SAG: sagittal.
表2  图像质量Rickett量表主观评分标准
Tab. 2  Subjective scoring criteria of Rickett scale of image quality
表3  两名影像科医师对传统TSE序列及DL-TSE序列图像主观评价评分的一致性分析
Tab. 3  Consistency analysis of the subjective evaluation scores of TSE sequence and DL-TSE sequence images by two diagnostic doctors
表4  传统TSE序列与DL-TSE序列组内相关系数
Tab. 4  Intra-group correlation coefficient between conventional TSE sequences and DL-TSE sequences
表5  传统TSE序列与DL-TSE序列主观评分比较
Tab. 5  Comparison of subjective scores of traditional TSE sequences and DL-TSE sequences

1.5 统计学分析

       采用SPSS 26.0软件(IBM SPSS Statistics for Windows)进行统计分析。采用Shapiro-Wilk检验评价定量数值和主观评分的正态分布性,服从正态分布的数据采用配对样本检验,以均数±标准差(x¯±s)表示;不服从正态分布的数据采用配对样本非参数检验(Wilcoxon符号轶检验),以中位数和四分位数MQ25,Q75)表示。对两名影像科技师/影像诊断科医生测得的SNR值、CNR值和主观评价评分的一致性进行组内相关系数分析,ICC值>0.75证明数据一致性较好。P<0.05为差异有统计学意义。

2 结果

2.1 入组患者一般资料

       本研究经排除标准排除4名患者,最终纳入60名患者,男30名,女30名,年龄35~69(53±10)岁。传统TSE序列扫描总时间为10分17秒,DL-TSE序列扫描总时间为6分31秒,与传统TSE序列相比,DL-TSE序列扫描时间缩短了36.6%。

2.2 一致性结果

       所有患者传统TSE序列及DL-TSE序列图像质量均可满足诊断需求,两名影像科医师主观评价中的四项评分指标观察者间一致性强,Kappa值范围为0.803~0.922,具有很好一致性(表3)。

       两名技师分别测得的传统TSE序列图像及DL-TSE序列图像的SNR和CNR具有很好的一致性,传统TSE序列SNR的ICC值范围为0.911~0.961,DL-TSE序列SNR的ICC值范围为0.918~0.972;传统TSE序列CNR的ICC值范围为0.893~0.987,DL-TSE序列CNR的ICC值范围为:0.914~0.981。ICC值均大于0.75证明数据一致性较好。采用两名影像科技师所测图像的SNR和CNR的平均值做统计学分析(表4)。

2.3 主观评价结果

       两名影像科医师对DL-TSE序列图像及传统TSE序列图像的主观评分进行比较,DL-TSE序列图像质量评分均显著高于传统TSE序列,两组图像间图像伪影、病灶轮廓清晰度及病灶结构清晰度测性评分差异具有统计学意义(P<0.001)(表5)。DL-TSE序列的图像质量和病变可检测性优于传统TSE序列,DL-TSE序列提供了更高的图像质量,提高了直肠疾病的诊断效率。

2.4 客观评价结果

       传统TSE序列中矢状T2WI、冠状T2WI、轴位小视野T2WI、轴位抑脂T2WI、轴位T1WI五组CNR、SNR分别与DL-TSE序列的五组两两比较发现,传统TSE序列和DL-TSE序列在SNR和CNR上的差异均具有统计学意义(P<0.001),详见表6。DL-TSE序列的SNR和CNR高于传统TSE序列,且差异有统计学意义,这表明DL Recon在缩短36.6%扫描时间的同时可增加SNR和CNR。

表6  传统TSE序列与DL-TSE序列图像客观评价比较
Tab. 6  Comparison of objective evaluation of traditional TSE sequences and DL-TSE sequences

3 讨论

       本项前瞻性研究是国内首次在临床环境中将DL Recon技术应用于直肠MRI,并比较了传统TSE序列和DL-TSE序列的扫描时间和图像质量。结果显示,DL-TSE序列不仅可以将扫描时间缩短了36.6%,并且明显改善了直肠MRI的图像质量,DL-TSE序列的SNR和CNR及整体图像质量评分均明显高于传统TSE序列。这表明应用DL-Recon技术提高TSE序列图像质量,同时缩短患者扫描时间成为可能。

3.1 DL Recon技术提高图像质量,缩短扫描时间

       传统TSE序列采集时间长且容易受到伪影影响,难以满足日益增长的临床需求,因此需要高质量且快速的直肠TSE序列。传统TSE序列需要较长的扫描时间,不仅增加了患者的不适感,也增加了产生运动伪影的可能性,这限制了其在临床中的应用。

       DL技术作为一种基于人工神经网络的机器学习方法,其核心优势在于能够通过大量的数据自动提取复杂的图像特征,从而实现高效的图像识别和分类。在医学影像领域中,DL技术已经被广泛应用于图像重建、病灶检测[19]、图像分割[20]、计算机辅助诊断[21]等多个任务中。

       由于人体结构的复杂性,临床对于各部位MRI检查的要求也各不相同。但通过大量的临床实践,使用DL Recon技术的MRI已初步应用在人体的不同部位,并有着不错的效果。HAHN等[22]对105名患者的肩关节进行了110次扫描,分别采用三个平面上标准2D TSE序列和三个平面上使用DL Recon加速的序列,结果显示使用DL Recon的序列提高了图像质量,扫描时间比传统TSE序列缩短67%,在保持图像质量和诊断性能方面节省了大量时间。KIDOH等[23]对15名志愿者的颅脑采用传统2D TSE序列及DL-TSE序列进行扫描,结果显示减少相位编码数或减少激励次数的情况下DL-TSE图像质量仍不逊于传统TSE序列图像,不仅能有效去除颅脑图像伪影,还能减少大量扫描时间。徐旭等[24]对60名患者在1.5 T MRI上采用传统TSE序列及DL-TSE序列进行上腹部MRI检查,结果显示DL-TSE序列在图像噪声、清晰度、对比度、总体图像质量上均优于传统TSE序列。JOHNSON等[25]对170名患者的膝关节进行扫描,并将DL Recon序列的图像与传统扫描序列图像进行比较分析,结果表明DL Recon的图像质量评分均超过传统扫描序列图像,并使得扫描时间减少了近一半,这与XIE等[26]的研究结果一致。此外DL Recon技术在脑[27]、腹部[28]、乳腺[29, 30, 31]等部位的MRI图像质量均具有良好的表现,有望打破MRI扫描时间与图像质量不能兼顾的局面[32, 33]。然而,目前仅有一项研究比较了超高分辨率采集和DL Recon 对T2WI序列及增强重建[螺旋桨技术(periodically rotated overlapping parallel lines with enhanced reconstruction, PROPELLER)]直肠MRI图像质量和诊断性能的影响[34],结果显示,应用DL Recon技术的T2WI及超高分辨率PROPELLER的图像质量均有显著提高,然而该研究没有对T1WI序列及扫描时间进行分析。本研究结果显示,DL-TSE图像在SNR和CNR上均优于传统TSE序列图像,并在提升图像质量的同时缩短了扫描时间,进一步减少了扫描过程中可能会出现的伪影,这与MATSUMOTO等的研究结果一致[35, 36, 37, 38]

       在本研究中,主观评价结果及客观评价结果显示DL-TSE各序列的评分均优于传统TSE序列,提高了图像重建的效率和精度,提示DL-TSE序列的稳定性及可重复性能满足临床诊断要求,为直肠癌的诊断与治疗提供了新的可能性。

3.2 局限性及展望

       本研究尚存在一些不足:第一,本研究仅探讨了DL-TSE序列提高直肠MRI图像质量的可能性,未将DL-TSE序列图像对疾病的诊断影响纳入研究范围;第二,本次研究单中心研究且样本量较小,后续研究中会继续扩充样本量或对DL-TSE序列是否可以提高其他部位图像质量做出探讨。

4 结论

       综上所述,本研究探讨了DL Recon对提高直肠癌MRI图像质量的可行性。DL Recon在直肠癌MRI TSE序列中的T2WI和T1WI上表现良好,与传统TSE序列相比,既能更快地获取图像,又能在减少扫描伪影的情况下保持高图像质量。在3.0 T MRI扫描中相较于传统的TSE序列可缩短36.6%的时间。

[1]
SIEGEL R L, MILLER K D, FUCHS H E, et al. Cancer statistics, 2022[J]. CA Cancer J Clin, 2022, 72(1): 7-33. DOI: 10.3322/caac.21708.
[2]
SIEGEL R L, MILLER K D, WAGLE N S, et al. Cancer statistics, 2023[J]. CA Cancer J Clin, 2023, 73(1): 17-48. DOI: 10.3322/caac.21763.
[3]
SIEGEL R L, GIAQUINTO A N, JEMAL A. Cancer statistics, 2024[J]. CA Cancer J Clin, 2024, 74(1): 12-49. DOI: 10.3322/caac.21820.
[4]
王劲松, 魏家燕, 彭敏. 2023年美国癌症统计报告和全球最新癌症统计数据解读及启示[J]. 实用肿瘤杂志, 2023, 38(6): 523-527. DOI: 10.13267/j.cnki.syzlzz.2023.083.
WANG J S, WEI J Y, PENG M. Interpretation and enlightenment of 2023 American cancer statistics report and the latest global cancer statistics[J]. J Pract Oncol, 2023, 38(6): 523-527. DOI: 10.13267/j.cnki.syzlzz.2023.083.
[5]
杨井飞, 陈英, 宋伟. MRI在中低位直肠癌患者经肛全直肠系膜切除术前分期评估中的应用价值[J]. 浙江创伤外科, 2023, 28(12): 2386-2389. DOI: 10.3969/j.issn.1009-7147.2023.12.057.
YANG J F, CHEN Y, SONG W. Application value of MRI in staging evaluation of patients with middle and low rectal cancer before transanal total mesorectal excision[J]. Zhejiang J Trauma Surg, 2023, 28(12): 2386-2389. DOI: 10.3969/j.issn.1009-7147.2023.12.057.
[6]
OGAWA M, URANO M, TAKAISHI T, et al. T-staging of rectal cancer: utility of single-shot turbo spin-echo diffusion-weighted imaging with T2-weighted images and fusion images[J/OL]. PLoS One, 2021, 16(4): e0249433 [2024-04-21]. https://pubmed.ncbi.nlm.nih.gov/33882087. DOI: 10.1371/journal.pone.0249433.
[7]
卫佳佳, 姬丽娟. 核磁共振(MRI)和CT评估直肠癌术前T分期的临床价值[J]. 中国CT和MRI杂志, 2017, 15(12): 113-116, 139. DOI: 10.3969/j.issn.1672-5131.2017.12.035.
WEI J J, JI L J. Clinical value of magnetic resonance imaging (MRI) and CT in evaluating the preoperative T stage of rectal cancer[J]. Chin J CT MRI, 2017, 15(12): 113-116, 139. DOI: 10.3969/j.issn.1672-5131.2017.12.035.
[8]
LV B H, CHENG X J, XIE Y Z, et al. Predictive value of lesion morphology in rectal cancer based on MRI before surgery[J/OL]. BMC Gastroenterol, 2023, 23(1): 318 [2024-04-21]. https://pubmed.ncbi.nlm.nih.gov/37726671/. DOI: 10.1186/s12876-023-02910-4.
[9]
陈楚瑶, 赵茜茜, 许乙凯. MRI观察壁外血管侵犯评估直肠癌新辅助治疗后根治性手术患者预后[J]. 中国医学影像技术, 2020, 36(12): 1848-1852. DOI: 10.13929/j.issn.1003-3289.2020.12.019.
CHEN C Y, ZHAO Q Q, XU Y K. MRI diagnosis of extramural vascular invasion for assessing prognosis of rectal cancer patients after neoadjuvant therapy and radical resection[J]. Chin J Med Imag Technol, 2020, 36(12): 1848-1852. DOI: 10.13929/j.issn.1003-3289.2020.12.019.
[10]
GÜRSES B, BÖGE M, ALTıNMAKAS E, et al. Multiparametric MRI in rectal cancer[J]. Diagn Interv Radiol, 2019, 25(3): 175-182. DOI: 10.5152/dir.2019.18189.
[11]
FERNANDES M C, GOLLUB M J, BROWN G. The importance of MRI for rectal cancer evaluation[J/OL]. Surg Oncol, 2022, 43: 101739 [2024-03-18]. https://pubmed.ncbi.nlm.nih.gov/35339339. DOI: 10.1016/j.suronc.2022.101739.
[12]
VAN DER POL C B, SHABANA W M, MCINNES M D, et al. High-resolution T2-weighted (T2W) oblique plane turbo spin-echo (TSE) MRI for rectal adenocarcinoma staging[J]. Clin Imaging, 2015, 39(4): 627-631. DOI: 10.1016/j.clinimag.2015.01.014.
[13]
韩峰, 徐亚明. 认识磁共振检查中图像伪影的常见类型及应对解决策略[J]. 中国医疗器械信息, 2019, 25(20): 22-23. DOI: 10.3969/j.issn.1006-6586.2019.20.011.
HAN F, XU Y M. Understanding the common types of image artifacts in magnetic resonance examination and their solutions[J]. China Med Device Inf, 2019, 25(20): 22-23. DOI: 10.3969/j.issn.1006-6586.2019.20.011.
[14]
FRITZ J, GUGGENBERGER R, DEL GRANDE F. Rapid musculoskeletal MRI in 2021: clinical application of advanced accelerated techniques[J]. AJR Am J Roentgenol, 2021, 216(3): 718-733. DOI: 10.2214/AJR.20.22902.
[15]
朱承刚. 654-2(山莨菪碱)+手法推压在治疗原发性肠套叠中的临床应用[J]. 中国实用医药, 2012, 7(30): 152. DOI: 10.3969/j.issn.1673-7555.2012.30.117.
ZHU C G. Clinical application of 654-2 (anisodamine) plus manual pushing and pressing in the treatment of primary intussusception[J]. China Pract Med, 2012, 7(30): 152. DOI: 10.3969/j.issn.1673-7555.2012.30.117.
[16]
ZHANG Y, ZOU J Y, WAN F, et al. Update on the sources, pharmacokinetics, pharmacological action, and clinical application of anisodamine[J/OL]. Biomedecine Pharmacother, 2023, 161: 114522 [2023-03-14]. https://pubmed.ncbi.nlm.nih.gov/37002581. DOI: 10.1016/j.biopha.2023.114522.
[17]
LUSTIG M, PAULY J M. SPIRiT: Iterative self-consistent parallel imaging reconstruction from arbitrary k-space[J]. Magn Reson Med, 2010, 64(2): 457-471. DOI: 10.1002/mrm.22428.
[18]
HAMMERNIK K, KLATZER T, KOBLER E, et al. Learning a variational network for reconstruction of accelerated MRI data[J]. Magn Reson Med, 2018, 79(6): 3055-3071. DOI: 10.1002/mrm.26977.
[19]
王进, 冯友红. 基于深度学习技术的医用影像检测方法研究[J]. 电子设计工程, 2021, 29(20): 169-173. DOI: 10.14022/j.issn1674-6236.2021.20.036.
WANG J, FENG Y H. Research on medical image detection method based on deep learning technology[J]. Electron Des Eng, 2021, 29(20): 169-173. DOI: 10.14022/j.issn1674-6236.2021.20.036.
[20]
刘彬. 基于深度学习的医学图像分割与应用研究[D]. 烟台: 烟台大学, 2023.
LIU B. Research on medical image segmentation and application based on deep learning[D].Yantai: Yantai University, 2023.
[21]
殷保才. 基于深度学习的医学影像辅助诊断技术研究[D]. 合肥: 中国科学技术大学, 2022.
YIN B C. Research on medical image-aided diagnosis technology based on deep learning[D].Hefei: University of Science and Technology of China, 2022.
[22]
HAHN S, YI J, LEE H J, et al. Image quality and diagnostic performance of accelerated shoulder MRI with deep learning-based reconstruction[J]. AJR Am J Roentgenol, 2022, 218(3): 506-516. DOI: 10.2214/AJR.21.26577.
[23]
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(3): 195-206. DOI: 10.2463/mrms.mp.2019-0018.
[24]
徐旭, 彭婉琳, 张金戈, 等. 基于人工智能的滤波和插值图像重建算法在腹部磁共振图像降噪中的价值[J]. 四川大学学报(医学版), 2021, 52(2): 293-299. DOI: 10.12182/20210360104.
XU X, PENG W L, ZHANG J G, et al. The application value of artificial intelligence-based filtering and interpolated image reconstruction algorithm in abdominal magnetic resonance image denoising[J]. J Sichuan Univ Med Sci, 2021, 52(2): 293-299. DOI: 10.12182/20210360104.
[25]
JOHNSON P M, LIN D J, ZBONTAR J, et al. Deep learning reconstruction enables prospectively accelerated clinical knee MRI[J/OL]. Radiology, 2023, 307(2): e220425 [2023-03-14]. https://pubmed.ncbi.nlm.nih.gov/36648347/. DOI: 10.1148/radiol.220425.
[26]
XIE Y X, TAO H Y, LI X W, et al. Prospective comparison of standard and deep learning-reconstructed turbo spin-echo MRI of the shoulder[J/OL]. Radiology, 2024, 310(1): e231405 [2024-01-09]. https://pubmed.ncbi.nlm.nih.gov/38193842. DOI: 10.1148/radiol.231405.
[27]
KHAN M K H, GUO W J, LIU J, et al. Machine learning and deep learning for brain tumor MRI image segmentation[J]. Exp Biol Med, 2023, 248(21): 1974-1992. DOI: 10.1177/15353702231214259.
[28]
OTTENS T, BARBIERI S, ORTON M R, et al. Deep learning DCE-MRI parameter estimation: application in pancreatic cancer[J/OL]. Med Image Anal, 2022, 80: 102512 [2024-03-01]. https://pubmed.ncbi.nlm.nih.gov/35709559/. DOI: 10.1016/j.media.2022.102512.
[29]
ZHAO X, BAI J W, GUO Q, et al. Clinical applications of deep learning in breast MRI[J/OL]. Biochim Biophys Acta Rev Cancer, 2023, 1878(2): 188864 [2023-09-03]. https://pubmed.ncbi.nlm.nih.gov/36822377/. DOI: 10.1016/j.bbcan.2023.188864.
[30]
SLANETZ P J. The potential of deep learning to revolutionize current breast MRI practice[J/OL]. Radiology, 2023, 306(3): e222527 [2023-11-15]. https://pubs.rsna.org/doi/10.1148/radiol.222527. DOI: 10.1148/radiol.222527.
[31]
BURGER B, BERNATHOVA M, SEEBÖCK P, et al. Deep learning for predicting future lesion emergence in high-risk breast MRI screening: a feasibility study[J/OL]. Eur Radiol Exp, 2023, 7(1): 32 [2024-03-01]. https://pubmed.ncbi.nlm.nih.gov/37280478/. DOI: 10.1186/s41747-023-00343-y.
[32]
HIGAKI T, NAKAMURA Y, TATSUGAMI F, et al. Improvement of image quality at CT and MRI using deep learning[J]. Jpn J Radiol, 2019, 37(1): 73-80. DOI: 10.1007/s11604-018-0796-2.
[33]
周楠, 花立春, 刘杰, 等. 深度学习重建法在MRI重建中的应用进展[J]. 中国医疗设备, 2023, 38(12): 165-169. DOI: 10.3969/j.issn.1674-1633.2023.12.029.
ZHOU N, HUA L C, LIU J, et al. Application of deep learning reconstruction in MRI reconstruction[J]. China Med Devices, 2023, 38(12): 165-169. DOI: 10.3969/j.issn.1674-1633.2023.12.029.
[34]
MATSUMOTO S, TSUBOYAMA T, ONISHI H, et al. Ultra-high-resolution T2-weighted PROPELLER MRI of the rectum with deep learning reconstruction: assessment of image quality and diagnostic performance[J]. Invest Radiol, 2024, 59(7): 479-488. DOI: 10.1097/RLI.0000000000001047.
[35]
KANIEWSKA M, DEININGER-CZERMAK E, GETZMANN J M, et al. Application of deep learning-based image reconstruction in MR imaging of the shoulder joint to improve image quality and reduce scan time[J]. Eur Radiol, 2023, 33(3): 1513-1525. DOI: 10.1007/s00330-022-09151-1.
[36]
HALLINAN J T P D. Deep learning for spine MRI: reducing time not quality[J/OL]. Radiology, 2023, 306(3): e222410 [2023-11-01]. https://pubmed.ncbi.nlm.nih.gov/36318033. DOI: 10.1148/radiol.222410.
[37]
PAPP D, CASTILLO T J M, WIELOPOLSKI P A, et al. Deep learning for improving ZTE MRI images in free breathing[J]. Magn Reson Imaging, 2023, 98: 97-104. DOI: 10.1016/j.mri.2023.01.019.
[38]
ZHANG C, SCHWARTZ M, KÜSTNER T, et al. Multiparametric functional MRI of the kidney: current state and future trends with deep learning approaches[J]. Rofo, 2022, 194(9): 983-992. DOI: 10.1055/a-1775-8633.

上一篇 基于DCE-MRI图像深度学习模型鉴别诊断乳腺良恶性肿瘤的价值分析
下一篇 基于深度学习重建扩散加权成像图像评估甲状腺相关性眼病的活动性
  
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