基于深度学习的快速MRCP成像在屏气困难患者检查中的应用价值

doi:10.16150/j.1671-2870.2025.06.008

摘要/Abstract

摘要：

目的： 比较基于快速深度学习（deep learning speed，DLS）的三维屏气（breathe hold，BH）磁共振胰胆管成像（magnetic resonance cholangiopancreatography ，MRCP）与常规的三维屏气磁共振胰胆管成像（3D BH MRCP）在无法配合屏气的情况下的图像质量。方法： 前瞻性纳入2025年1月至3月在上海交通大学医学院附属瑞金医院50例屏气配合不佳MRCP检查的患者，先后进行BH 3D DLS MRCP、BH 3D MRCP序列的扫描。比较2个序列的扫描时间，客观评价2组图像的信噪比（signal to noise ratio， SNR）、对比噪声比（contrast to noise ratio，CNR），主观评价2组图像的整体图像质量、背景抑制、胰胆管可视化质量。结果： 3D BH DLS MRCP组平均扫描时间为11 s，3D BH MRCP组平均扫描时间为17 s，3D BH DLS MRCP扫描时间缩短（P<0.05）。客观评价显示3D BH DLS MRCP的SNR（23.76±9.23）和CNR（28.89±11.49）均好于3D BH MRCP[（17.15±6.85）、（22.03±8.79）]（P<0.05）；主观评价图像分析显示，3D BH DLS MRCP在图像整体质量评分[（3.80±0.32）分]、背景抑制[（3.64±0.40）分]及显示胰胆管各段[（3.53±0.37）分）]均有高于3D BH MRCP的图像整体质量[（3.10±0.44）分]、背景抑制[（3.33±0.51分]）及显示胰胆管各段[（3.08±0.56）分]的趋势，但差异尚无统计学意义。3D BH DLS MRCP组50例均满足诊断要求，3D BH MRCP组有5例不满足诊断要求，2组间差异有统计学意义（P<0.05）。结论： 在无法屏气或屏气不佳的情况下，应用3D BH DLS MRCP检查可显著缩短检查时间，图像质量有保证，符合临床诊断需求。

关键词: 磁共振胰胆管成像, 快速深度学习, 屏气

Abstract:

Objective To compare the image quality of rapid deep learning (RDL)-based three-dimensional (3D) breath holding (BH) magnetic resonance cholangiopancreatography (MRCP) and that of conventional 3D BH MRCP in patients unable to cooperate with breath holding. Methods A total of 50 patients with poor breath holding cooperation who underwent MRCP examination at Ruijin Hospital affiliated with Shanghai Jiao Tong University School of Medicine from January to March 2025 were prospectively enrolled for scanning with 3D BH DLS MRCP(DLS group) and BH 3D MRCP(MRCP group) sequences successively. The scanning duration of the two sequences were compared. Objective analysis including the signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) of the images, as well as subjective analysis including the overall image quality, background suppression, and visualization of the pancreatic and biliary ducts in both groups were compared. Results The averge scanning duration was 11 s for the 3D BH DLS MRCP group and 17 s for the 3D BH MRCP group, indicating a significant reduction in scanning time for 3D BH DLS MRCP (P<0.05). Objective analysis revealed that both the SNR (23.76±9.23) and CNR (28.89±11.49) of 3D BH DLS MRCP were higher than those of 3D BH MRCP [(17.15±6.85), (22.03±8.79)] (P<0.05). Subjective image analysis showed that 3D BH DLS MRCP outperformed 3D BH MRCP in overall image quality scores (3.80±0.32 vs 3.10±0.44), background suppression (3.64±0.40 vs 3.33±0.51), and visualization of pancreaticobiliary duct segments (3.53±0.37 vs 3.08±0.56) with an upward trend(P>0.05). Results of all 50 cases in the 3D BH DLS MRCP sequence met the diagnostic requirements, while of 5 cases in the 3D BH MRCP sequence did not, with a statistically significant difference between the two sequences (P<0.05). Conclusions In cases where breath holding is impossible or poorly performed, the use of 3D BH DLS MRCP could significantly shortens examination time, with image quality of the same level as BH 3D MRCP, and meets clinical diagnostic requirements.

Key words: Magnetic resonance cholangiopancreatography, Rapid deep learning, Breath holding

中图分类号:

R445.2

孔德艳, 高祎阳, 张曹阳, 董海鹏, 严福华, 原子扬, 代建琨, 刘方韬. 基于深度学习的快速MRCP成像在屏气困难患者检查中的应用价值[J]. 诊断学理论与实践, 2025, 24(06): 628-633.

KONG Deyan, GAO Yiyang, Zhang Caoyang, DONG Haipeng, YAN Fuhua, YUAN Ziyang, DAI Jiankun, LIU Fangtao. Application value of deep learning-based rapid MRCP imaging in examination of patients with breath holding difficulties[J]. Journal of Diagnostics Concepts & Practice, 2025, 24(06): 628-633.

图/表 6

表1

表2

图1

图2

图3

图4

参考文献 24

[1]	王嘉琛, 何思怡, 曹梦迪, 等. 1990—2021年中国人群5种常见消化系统恶性肿瘤疾病负担变化趋势分析[J]. 中华消化外科杂志, 2025, 24(2):213-222.
	WANG J C, HE S Y, CAO M D, et al. Analysis of the change trend in the burden of 5 common malignant tumors of digestive system in the Chinese population from 1990 to 2021[J]. Chin J Dig Surg, 2025, 24(2):213-222.
[2]	程琳, 余永强, 王成林, 等. 胰胆管汇合MRCP解剖与胰胆系疾病关系[J]. 中国CT和MRI杂志, 2012, 10(1):50-53.
	CHENG L, YU Y Q, WANG C L, et al. MRCP study of pancreaticobiliary maljunction and pancreaticobiliary di-seases[J]. Chin J CT MRI, 2012, 10(1): 50-53.
[3]	CAI L, YEH B M, WESTPHALEN A C, et al. 3D T2-weighted and Gd-EOB-DTPA-enhanced 3D T1-weighted MR cholangiography for evaluation of biliary anatomy in living liver donors[J]. Abdom Radiol (NY), 2017, 42(3):842-850. doi: 10.1007/s00261-016-0936-z pmid: 27714420
[4]	SANDINO C M, CHENG J Y, CHEN F, et al. Compressed sensing: from research to clinical practice with deep neural networks[J]. IEEE Signal Process Mag, 2020, 37(1):111-127.
[5]	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 URL
[6]	CHANDRA S S, BRAN LORENZANA M, LIU X, et al. Deep learning in magnetic resonance image reconstruction[J]. J Med Imaging Radiat Oncol, 2021, 65(5):564-577. doi: 10.1111/ara.v65.5 URL
[7]	UEDA T, OHNO Y, YAMAMOTO K, et al. Deep Lear-ning Reconstruction of Diffusion-weighted MRI Improves Image Quality for Prostatic Imaging[J]. Radiology, 2022, 303(2):373-381. doi: 10.1148/radiol.204097 URL
[8]	TAJIMA T, AKAI H, SUGAWARA H, et al. Feasibility of accelerated whole-body diffusion-weighted imaging using a deep learning-based noise-reduction technique in patients with prostate cancer[J]. Magn Reson Imaging, 2022, 92:169-179. doi: 10.1016/j.mri.2022.06.014 URL
[9]	JOHNSON P M, LIN D J, ZBONTAR J, et al. Deep learning reconstruction enables prospectively accelerated clinical knee MRI[J]. Radiology, 2023, 307(2):e220425.
[10]	ALMANSOUR H, HERRMANN J, GASSENMAIER S, et al. Deep learning reconstruction for accelerated spine MRI: Prospective analysis of interchangeability[J]. Radio-logy, 2023, 306(3):e212922.
[11]	MALKIEL I, AHN S, SLAVENS Z, et al. Densely connected iterative network for sparse MRI reconstruction[C]. Proceedings of the International Society for Magnetic Resonance in Medicine (ISMRM), Annual Meeting, 2017.
[12]	徐卓凡, 靳琪奥, 王开宇, 等. 基于CT检查的集成深度学习模型对肝门静脉定性与定量分型研究[J]. 中华消化外科杂志, 2024, 23(7):976-983.
	XU Z F, JIN Q A, WANG K Y, et al. CT-based integrated deep learning model for qualitative and quantitative research of hepatic portal vein[J]. Chin J Dig Surg, 2024, 23(7):976-983.
[13]	陶海粟, 黎柏宏, 曾小军, 等. 基于深度学习构建微创肝切除术关键解剖结构识别模型的应用价值[J]. 中华消化外科杂志, 2024, 23(4):590-595.
	TAO H S, LI B H, ZENG X J, et al. Application value of major anatomical structure recognition model of minimally invasive liver resection based on deep learning[J]. Chin J Dig Surg, 2024, 23(4):590-595.
[14]	KROMREY M L, FUNAYAMA S, TAMADA D, et al. Clinical evaluation of respiratory-triggered 3D MRCP with navigator echoes compared to breath-hold acquisition using compressed sensing and/or parallel imaging[J]. Magn Reson Med Sci, 2020, 19(4):318-323. doi: 10.2463/mrms.mp-2019-0122 URL
[15]	张小斌, 李宁, 陈亚明. MRCP 诊断不同直径、不同部位胆总管结石的价值[J]. 中国 CT 和 MRI 杂志, 2023, 21(4):110-111.
	ZHANG X B, LI N, CHEN Y M. The value of MRCP in the diagnosis of common bile duct stones of different diameters and locations[J]. Chin J CT MRI, 2023, 21(4):110-111.
[16]	王淳正, 许来艳, 侯莉莉. EUS检查与MRCP成像对IPMN良恶性鉴别诊断效能对比[J]. 中国CT和MRI杂志, 2022, 20(5):139-141.
	WANG C Z, XU L Y, HOU L L. Comparison on efficiency between EUS and MRCP imaging in the differential diagnosis of benign and malignant IPMN[J]. Chin J CT MRI, 2022, 20(5):139-141.
[17]	孟菲, 于霞. 3D MRCP及MIP图像与轴位T2WI结合诊断胆系结石的价值[J]. 武警医学, 2015, 26(7):656-658.
	MENG F, YU X. Diagnostic value of combining 3D MRCP/MIP images with axial T2-weighted imaging for biliary calculi[J]. Med J Chin PAPF, 2015, 26(7):656-658.
[18]	宋斌, 赵厚亮, 叶莉. 三维磁共振胰胆管成像诊断小儿胆道闭锁价值分析[J]. 实用肝脏病杂志, 2021, 24(2):288-291.
	SONG B, ZHAO H L, YE L. Application of three-dimensional magnetic resonance cholangiopancreatography in diagnosis of biliary atresia in children[J]. J Prac Hepatol, 2021, 24(2): 288-291.
[19]	王宏光, 罗漫. 肝门部胆管癌的术前评估和术中导航研究进展[J]. 中华消化外科杂志,2024,23(7)：906-911.
	WANG HG, LUO M. Research advance in preoperative evaluation and intraoperative navigation for hilar cholangio-carcinoma[J]. Chin J Dig Surg,2024,23(7)：906-911.
[20]	MALKIEL I, AHN S, SLAVENS Z, TAVIANI V, HARDY C J. Densely connected iterative network for sparse MRI reconstruction[C]. Proceedings of the International Society for Magnetic Resonance in Medicine (ISMRM), Annual Meeting, 2017.
[21]	GAO HUANG, ZHUANG LIU, LAURENS V D M, et al. Densely connected convolutional networks[C]. 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, USA,2017: 2261-2269.
[22]	AGGARWAL H K, MANI M P, JACOB M. MoDL: Model-based deep learning architecture for inverse problems[J]. IEEE Trans Med Imaging, 2019, 38(2):394-405. doi: 10.1109/TMI.2018.2865356 URL
[23]	AHN S, MENINI A, MCKINNON G, et al. Contrast-weighted SSIM loss function for deep learning-based undersampled MRI reconstruction[C]. (ISMRM) Annual Meeting, 2020.
[24]	王璇, 王皓, 万云天, 等. 肩关节加速MRI应用深度学习重建算法的可行性与临床价值[J]. 中国临床研究, 2024, 37(8):1238-1243.
	WANG X, WANG H, WAN YT, et al. Feasibility and clinical value of deep learning reconstruction in accele-rated MRI of shoulder[J]. Chin J Clin Res, 2024, 37(8):1238-1243.

Group	Overall image quality	Background suppression	Pancreaticobiliary duct visualization
3D BH DLS MRCP	3.80$\pm $0.32	3.64$\pm $0.40	3.53$\pm $0.37
3D BH MRCP	3.10$\pm $0.44	3.33$\pm $0.51	3.08$\pm $0.56
P value	0.450	0.117	0.263