基于DBCNet的TOF-MRA中脑动脉树区域自动分割方法

doi:10.11938/cjmr20223046

波谱学杂志 ›› 2023, Vol. 40 ›› Issue (3): 320-331.doi: 10.11938/cjmr20223046

基于DBCNet的TOF-MRA中脑动脉树区域自动分割方法

张嘉骏¹,鲁宇澄²,鲍奕仿²,李郁欣²,耿辰^3,^4,^#(),胡伏原^1,^§(),戴亚康^1,^3,^*()

1.苏州科技大学电子与信息工程学院，江苏苏州 215009
2.复旦大学附属华山医院放射科，上海 200040
3.中国科学院苏州生物医学工程技术研究所，江苏苏州 215163
4.济南国科医工科技发展有限公司，山东济南 250000

收稿日期:2022-12-16 出版日期:2023-09-05 在线发表日期:2023-03-03
通讯作者: *Tel: 15850168495, E-mail: daiyk@sibet.ac.cn;§Tel: 15062382549, E-mail: fuyuanhu@usts.edu.cn;#Tel: 18662520228, E-mail: gengc@sibet.ac.cn.
基金资助:
国家自然科学基金资助项目(81971685);山东省自然基金资助项目(ZR2022QF093);江苏省重点研发计划(BE2022049-2);苏州市科技计划(SS202072);浙江省重点研发计划(2020ZJZC03)

An Automatic Segmentation Method of Cerebral Arterial Tree in TOF-MRA Based on DBCNet

ZHANG Jiajun¹,LU Yucheng²,BAO Yifang²,LI Yuxin²,GENG Chen^3,^4,^#(),HU Fuyuan^1,^§(),DAI Yakang^1,^3,^*()

1. School of Electronics and Information Engineering, Suzhou University of Science and Technology, Suzhou 215009, China
2. Department of Radiology, Huashan Hospital, Fudan University, Shanghai 200040, China
3. Suzhou Institute of Biomedical Engineering and Technology, Chinese Academy of Sciences, Suzhou 215163, China
4. Jinan Guoke Medical Industry Technology Development Co., Jinan 250000, China

Received:2022-12-16 Published:2023-09-05 Online:2023-03-03
Contact: *Tel: 15850168495, E-mail: daiyk@sibet.ac.cn;§Tel: 15062382549, E-mail: fuyuanhu@usts.edu.cn;#Tel: 18662520228, E-mail: gengc@sibet.ac.cn.

摘要/Abstract

摘要：

从脑部医学影像中划分动脉树区域是诊断和评估许多脑血管疾病的早期步骤．现有的区域分割方法多依赖人工辅助，本文中提出了一种基于双分支连通网络（dual branch connected network，DBCNet）的脑动脉树自动分区方法，可以将时间飞跃磁共振血管造影（time of flight-magnetic resonance angiography，TOF-MRA）中的动脉树分割为6个主要区域．DBCNet中引入了分支特征解耦模块和Swin Transformer机制的全局与局部特征融合模块，训练采用先定位后分割的两步训练策略．本研究使用了111例TOF-MRA数据，其中81例作为训练集，20例作为验证集，10例作为测试集，模型在测试集上的平均Dice系数为74.72%，95%豪斯多夫距离（HD95）为3.89 mm．和其他先进分割网络相比较，该网络能更准确地分割出各个主要区域，并具有一定的鲁棒性．

关键词: 脑动脉树, 时间飞跃磁共振血管造影（TOF-MRA）, 深度学习, 分支连通网络, 自动分割

Abstract:

Arterial tree region segmentation from medical images of the brain is an early step in the diagnosis and evaluation of many cerebrovascular diseases. Most of the existing region segmentation methods rely on manual assistance. In this paper, we propose an automatic brain arterial tree partitioning method based on a dual branch connected network (DBCNet), which can partition the arterial tree in time of flight-magnetic resonance angiography (TOF-MRA) into six main regions. The branch feature decoupling module and the global and local feature fusion module based on Swin Transformer mechanism were used for DBCNet. The two-step training strategy of localization followed by segmentation was used for training. In this study, 111 cases of TOF-MRA data were used, of which 81 cases as the training set, 20 cases as the validation set, and 10 cases as the test set. The average Dice coefficient of the model on the test set was 74.72% and 95% Haus dorff distance (HD95) was 3.89 mm. Compared with other advanced segmentation networks, the network reported in this paper can segment each major region more accurately with robustness.

Key words: cerebral arterial tree, time of flight-magnetic resonance angiography (TOF-MRA), deep learning, dual branch connected network, automatic segmentation

中图分类号:

TP394.1

张嘉骏, 鲁宇澄, 鲍奕仿, 李郁欣, 耿辰, 胡伏原, 戴亚康. 基于DBCNet的TOF-MRA中脑动脉树区域自动分割方法[J]. 波谱学杂志, 2023, 40(3): 320-331.

ZHANG Jiajun, LU Yucheng, BAO Yifang, LI Yuxin, GENG Chen, HU Fuyuan, DAI Yakang. An Automatic Segmentation Method of Cerebral Arterial Tree in TOF-MRA Based on DBCNet[J]. Chinese Journal of Magnetic Resonance, 2023, 40(3): 320-331.

图/表 14

图1

表1

图2

图3

图4

图5

图6

表2

图7

表3

图8

图9

表4

图10

参考文献 29

[1]	GERI O, SHIRAN S I, ROTH J, et al. Vascular territorial segmentation and volumetric blood flow measurement using dynamic contrast enhanced magnetic resonance angiography of the brain[J]. J Cereb Blood Flow Metab, 2017, 37(10): 3446-3456. doi: 10.1177/0271678X17702394
[2]	TAHER F, PRAKASH N. Automatic cerebrovascular segmentation methods-a review[J]. IAES International Journal of Artificial Intelligence, 2021, 10(3): 576.
[3]	GAO X, UCHIYAMA Y, ZHOU X, et al. A fast and fully automatic method for cerebrovascular segmentation on time-of-flight (TOF) MRA image[J]. J Digit Imaging, 2011, 24(4): 609-625. doi: 10.1007/s10278-010-9326-1 pmid: 20824304
[4]	CHEN M, GENG C, LI Y X, et al. Automatic detection for cerebral aneurysms in TOF-MRA images based on fuzzy label and deep learning[J]. Chinese J Magn Reson, 2022, 39(3): 267-277.
	陈萌, 耿辰, 李郁欣, 等. 基于模糊标签和深度学习的TOF-MRA影像脑动脉瘤自动检测[J]. 波谱学杂志, 2022, 39(3): 267-277.
[5]	REN Y, CHEN G Z, LIU Z, et al. Reproducibility of image-based computational models of intracranial aneurysm: a comparison between 3D rotational angiography, CT angiography and MR angiography[J]. Biomed Eng Online. 2016, 15: 50. doi: 10.1186/s12938-016-0163-4 pmid: 27150439
[6]	MU N, LYU Z, REZAEITALESHMAHALLEH M, et al. An attention residual U-Net with differential preprocessing and geometric postprocessing: Learning how to segment vasculature including intracranial aneurysms[J]. Med Image Anal, 2023, 84: 102697. doi: 10.1016/j.media.2022.102697
[7]	LI Y, NI J, ELAZAB A, et al. Multiple self-attention network for intracranial vessel segmentation[C]// International Joint Conference on Neural Networks, Online: IEEE, 2021: 1-8.
[8]	BIZJAK Ž, CHIEN A, BURNIK I, et al. Novel dataset and evaluation of state-of-the-art vessel segmentation methods[J]. SPIE, 2022, 12032, 120322x.
[9]	XIA L k, ZHANG H, WU Y, et al. 3D vessel-like structure segmentation in medical images by an edge-reinforced network[J]. Med Image Anal, 2022, 82: 102581. doi: 10.1016/j.media.2022.102581
[10]	JONES J D, CASTANHO P, BAZIRA P, et al. Anatomical variations of the circle of Willis and their prevalence, with a focus on the posterior communicating artery: A literature review and meta-analysis[J]. Clin Anat, 2021, 34(7): 978-990. doi: 10.1002/ca.v34.7
[11]	TAKEMURA A, SUZUKI M, HARAUCHI H, et al. Automatic segmentation method which divides a cerebral artery tree in time-of-flight MR-angiography into artery segments[J]. P Soc Photo Opt Instrum Eng, 2006, 6144: 1098-1106.
[12]	NOWINSKI W L, VOLKAU I, MARCHENKO Y, et al. A 3D model of human cerebrovasculature derived from 3T magnetic resonance angiography[J]. Neuroinformatics, 2009, 7(1): 23-36. doi: 10.1007/s12021-008-9028-8 pmid: 19016001
[13]	CHEN L, MOSSA-BASHA M, SUN J, et al. Quantification of morphometry and intensity features of intracranial arteries from 3D TOF MRA using the intracranial artery feature extraction (iCafe): A reproducibility study[J]. Magn Reson Imaging, 2019, 57: 293-302. doi: S0730-725X(18)30538-1 pmid: 30580079
[14]	CHEN L, SUN J, HIPPE D S, et al. Quantitative assessment of the intracranial vasculature in an older adult population using iCafe[J]. Neurobiology of Aging, 2019, 79: 59-65. doi: S0197-4580(19)30077-6 pmid: 31026623
[15]	LIU L L, CHENG J H, QUAN Q, et al. A survey on U-shaped networks in medical image segmentations[J]. Neurocomputing, 2020, 409: 244-258. doi: 10.1016/j.neucom.2020.05.070
[16]	QIU Y, NIE S D, WEI L. Segmentation of breast tumors based on fully convolutional network and dynamic contrast enhanced magnetic resonance image[J]. Chinese J Magn Reson, 2022, 39(2): 196-207.
	邱玥, 聂生东, 魏珑. 基于全卷积网络的乳腺肿瘤动态增强磁共振图像分割[J]. 波谱学杂志, 2022, 39(2): 196-207.
[17]	FAN D P, JI G P, SUN G, et al. Camouflaged object detection[C]// Computer Vision and Pattern Recognition, 2020: 2777-2787
[18]	YANG Z, SOLTANIAN-ZADEH S, FARSIU S. BiconNet: An edge-preserved connectivity-based approach for salient object detection[J]. Pattern Recogn, 2022, 121: 108231. doi: 10.1016/j.patcog.2021.108231
[19]	LIU Z, LIN Y T, CAO Y, et al. Swin transformer: Hierarchical vision transformer using shifted windows[C]// International Conference on Computer Vision, China:IEEE, 2021: 10012-10022.
[20]	YUSHKEVICH P, PIVEN J, HAZLETT H, et al. User-guided 3D active contour segmentation of anatomical structures: Significantly improved efficiency and reliability[J]. Neuroimage. 2006, 31(3): 1116-28. doi: 10.1016/j.neuroimage.2006.01.015 pmid: 16545965
[21]	INCI S, ERBENGI A, ÖZGEN T. Aneurysms of the distal anterior cerebral artery: report of 14 cases and a review of the literature[J]. Surg Neurol, 1998, 50(2): 130-140. pmid: 9701118
[22]	CANNY J. A computational approach to edge detection[J]. IEEE Trans Pattern Anal Mach Intell, 1986, 8(6): 679-98.
[23]	HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]// Computer Vision and Pattern Recognition, USA: IEEE, 2016: 770-778.
[24]	WOO S H, PARK J, LEE J Y, et al. CBAM: Convolutional block attention module[J]. Computer Vision, 2018, 11211: 3-19.
[25]	YEUNG M, SALA E, SCHÖNLIEB C B, et al. Unified focal loss: Generalising dice and cross entropy-based losses to handle class imbalanced medical image segmentation[J]. Comput Med Imag Grap, 2022, 95: 102026. doi: 10.1016/j.compmedimag.2021.102026
[26]	ISENSEE F, JAEGER P F, KOHL S A A, et al. nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation[J]. Nat Methods, 2021, 18(2): 203-211. doi: 10.1038/s41592-020-01008-z pmid: 33288961
[27]	ISENSEE F, KICKINGEREDER P, WICK W, et al. Brain tumor segmentation and radiomics survival prediction: Contribution to the BRATS 2017 Challenge[J]. Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries, 2017, 10670: 287-297.
[28]	MILLETARI F, NAVAB N, AHMADI S A, et al. V-Net: Fully convolutional neural networks for volumetric medical image segmentation[C]// International Conference on 3d Vision, 2016: 565-571.
[29]	ÇIÇEK Ö, ABDULKADIR A, LIENKAMP S S. 3D U-Net: Learning dense volumetric segmentation from sparse annotation[C]// Medical Image Computing and Computer-Assisted Intervention, 2016: 424-432.

血管疾病	重复时间	回波时间	视野百分比相位	采集矩阵	层内分辨率	图像层数	反转角	层厚
健康人	3.4 ms	25 ms	88%	320×192	0.43 mm×0.43 mm	128	20°	1.4 mm
动脉瘤患者	5.7 ms	25 ms	88%	320×256	0.21 mm×0.21 mm	240	20°	1.2/1.4 mm

	ACA	BA	ICA	MCA	PCA	VA	平均
Dice/%	86.32±4.59	81.56±3.54	92.52±1.25	86.53±3.44	81.66±3.15	82.92±6.28	74.72±3.36
HD95/mm	3.30±2.32	4.94±2.97	1.27±0.87	3.64±2.41	5.16±3.02	5.05±5.85	3.89±1.30

		ACA	BA	ICA	MCA	PCA	VA	平均
nnUNet	Dice/%	52.18±2.29	0	80.03±5.93	57.00±2.15	45.00±15.81	0	26.78±3.91
nnUNet	HD95/mm	31.74±8.50	30	8.29±11.01	21.91±14.10	27.15±4.08	30	24.84±3.94
Modified UNet	Dice/%	77.89±5.14	81.01±7.22	89.49±2.67	76.26±2.94	0	0	49.90±3.59
Modified UNet	HD95/mm	6.48±4.48	8.49±14.01	4.95±10.67	7.99±3.32	30	30	14.65±2.69
VNet	Dice/%	58.70±22.74	73.56±7.79	89.46±3.81	70.86±4.16	72.97±5.66	70.04±14.56	54.56±8.59
VNet	HD95/mm	13.56±8.43	12.59±14.59	5.16±11.03	15.76±4.84	22.81±6.57	14.33±6.69	14.04±8.69
DBCNet	Dice/%	86.32±4.59	81.56±3.54	92.52±1.25	86.53±3.44	81.66±3.15	82.92±6.28	74.72±3.36
DBCNet	HD95/mm	3.30±2.32	4.94±2.97	1.27±0.87	3.64±2.41	5.16±3.02	5.05±5.85	3.89±1.30

模型	消融策略	Dice/%	HD95/mm
DBCNet	无	87.36	1.47
	不使用交换权重训练策略	82.16 (↓5.20)	1.84 (↑0.37)
	双分支均使用实例标注	84.33 (↓3.03)	1.61 (↑0.14)
	不使用BiA模块	84.53 (↓2.83)	1.90 (↑0.43)
	不使用SC模块	84.62 (↓2.74)	2.00 (↑0.53)

基于DBCNet的TOF-MRA中脑动脉树区域自动分割方法

An Automatic Segmentation Method of Cerebral Arterial Tree in TOF-MRA Based on DBCNet

RichHTML

PDF

评审附件

可视化

摘要/Abstract

引用本文

使用本文

图/表 14

参考文献 29

相关文章 15

编辑推荐

Metrics

本文评价

[1]	李奕洁, 杨馨雨, 杨晓梅. 融合注意力机制的多尺度残差Unet的磁共振图像重建[J]. 波谱学杂志, 2023, 40(3): 307-319.
[2]	陆琪琪, 连梓锋, 李嘉龙, 斯文彬, 麦兆华, 冯衍秋. 基于自监督网络的肝脏磁共振R₂^*参数图像重建[J]. 波谱学杂志, 2023, 40(3): 258-269.
[3]	田慧, 武杰, 边云, 张志伟, 邵成伟. 基于DenseNet结合迁移学习的胰腺囊性肿瘤分类方法[J]. 波谱学杂志, 2023, 40(3): 270-279.
[4]	黄敏,李思怡,陈军波,周到. 磁共振指纹成像技术及临床应用的进展[J]. 波谱学杂志, 2023, 40(2): 207-219.
[5]	钱程一,王远军. 基于深度学习的阿尔兹海默症影像学分类研究进展[J]. 波谱学杂志, 2023, 40(2): 220-238.
[6]	施伟成,金朝阳,叶铮. 基于PCAU-Net的快速多通道磁共振成像方法[J]. 波谱学杂志, 2023, 40(1): 39-51.
[7]	周勤, 王远军. 基于变分推断的磁共振图像群组配准[J]. 波谱学杂志, 2022, 39(3): 291-302.
[8]	常晓,蔡昕,杨光,聂生东. 生成对抗网络在医学图像转换领域的应用[J]. 波谱学杂志, 2022, 39(3): 366-380.
[9]	王颖珊, 邓奥琦, 毛瑾玲, 朱中旗, 石洁, 杨光, 马伟伟, 路青, 汪红志. 基于3D VNetTrans的膝关节滑膜磁共振图像自动分割[J]. 波谱学杂志, 2022, 39(3): 303-315.
[10]	陈萌, 耿辰, 李郁欣, 耿道颖, 鲍奕仿, 戴亚康. 基于模糊标签和深度学习的TOF-MRA影像脑动脉瘤自动检测[J]. 波谱学杂志, 2022, 39(3): 267-277.
[11]	王振宇, 王颖珊, 毛瑾玲, 马伟伟, 路青, 石洁, 汪红志. 基于Dense-UNet++的关节滑膜磁共振图像分割[J]. 波谱学杂志, 2022, 39(2): 208-219.
[12]	霍璐,胡晓欣,肖勤,顾雅佳,褚旭,姜娈. 基于nnU-Net的乳腺DCE-MR图像中乳房和腺体自动分割[J]. 波谱学杂志, 2021, 38(3): 367-380.
[13]	刘鹏, 钟玉敏, 王丽嘉. 基于密集多尺度U-net网络的电影心脏磁共振图像右心室自动分割[J]. 波谱学杂志, 2020, 37(4): 456-468.
[14]	赵尚义, 王远军. 基于磁共振图像和改进的UNet++模型区分阿尔茨海默症患者和健康人群[J]. 波谱学杂志, 2020, 37(3): 321-331.
[15]	宫进昌, 王宇, 王远军. 结合小波融合和深度学习的脑胶质瘤自动分割[J]. 波谱学杂志, 2020, 37(2): 131-143.