《计算机应用》唯一官方网站 ›› 2025, Vol. 45 ›› Issue (6): 2007-2015.DOI: 10.11772/j.issn.1001-9081.2024060853
• 多媒体计算与计算机仿真 • 上一篇
陈盈涛1, 方康康1, 张金敖1, 梁浩然1, 郭焕斌2, 邱兆文1()
收稿日期:
2024-06-24
修回日期:
2024-09-27
接受日期:
2024-10-09
发布日期:
2024-10-29
出版日期:
2025-06-10
通讯作者:
邱兆文
作者简介:
陈盈涛(1999—),男,浙江临海人,硕士研究生,主要研究方向:医学影像分析基金资助:
Yingtao CHEN1, Kangkang FANG1, Jin’ao ZHANG1, Haoran LIANG1, Huanbin GUO2, Zhaowen QIU1()
Received:
2024-06-24
Revised:
2024-09-27
Accepted:
2024-10-09
Online:
2024-10-29
Published:
2025-06-10
Contact:
Zhaowen QIU
About author:
CHEN Yingtao, born in 1999, M. S. candidate. His research interests include medical image analysis.Supported by:
摘要:
冠状动脉血管形态结构的复杂性以及计算机断层扫描(CT)血管造影(CTA)图像的采集条件差异会导致图像灰度分布不均匀、运动伪影和噪声等图像质量问题,进而产生冠状动脉结构分割时的漏判和误判问题。因此,提出一种基于多尺度空间特征的冠状动脉CTA图像分割网络——三维多尺度并行网络(MSP-Net)。首先,针对冠状动脉的空间跨度大和局部占比小的特点,采用多尺度网络分别提取冠状动脉CTA图像的全局特征和局部特征并融合它们,保证冠状动脉结构特征的完整提取;其次,冠状动脉重建采用由粗到细的思想,增强图像特征的冗余性,确保冠状动脉的边界分明,再利用不同尺度特征度融合方法重建冠状动脉结构,以提高分割结果的准确率,减少漏判和误判;最后,为了加快网络的训练,采用深监督策略,在不同网络层级上引入监督信号,从而提高训练效率。实验结果表明,所提网络在冠状动脉自动分割任务中的平均Dice相似系数(DSC)达到87.16%,比nnU-Net和Swin UNETR (Swin UNEt TRansformers)分别提高了4.04和2.31个百分点,而平均95%豪斯多夫距离(HD95)达到3.69 mm,比nnU-Net和Swin UNETR分别降低了14.43 mm和13.75 mm。可见,所提网络能有效提高冠状动脉结构的分割精度,有助于临床医生更准确地了解患者冠状动脉结构,从而更有效地评估病情。
中图分类号:
陈盈涛, 方康康, 张金敖, 梁浩然, 郭焕斌, 邱兆文. 基于多尺度空间特征的冠状动脉CT血管造影图像分割网络[J]. 计算机应用, 2025, 45(6): 2007-2015.
Yingtao CHEN, Kangkang FANG, Jin’ao ZHANG, Haoran LIANG, Huanbin GUO, Zhaowen QIU. Segmentation network of coronary artery structure from CT angiography images based on multi-scale spatial features[J]. Journal of Computer Applications, 2025, 45(6): 2007-2015.
层序号 | 编码器输出特征图尺寸 | 通道数 | 解码器输出特征图尺寸 | 预测层输出特征图尺寸 | 损失函数权重 |
---|---|---|---|---|---|
1 | (B,32,96,160,160) | 32 | (B,32,96,160,160) | (B,C,96,160,160) | 1 |
2 | (B,64,48,80,80) | 64 | (B,64,48,80,80) | (B,C,48,80,80) | 1/2 |
3 | (B,128,24,40,40) | 128 | (B,128,24,40,40) | (B,C,24,40,40) | 1/4 |
4 | (B,256,12,20,20) | 256 | (B,256,12,20,20) | (B,C,12,20,20) | 1/8 |
5 | (B,320,6,10,10) | 320 | (B,320,6,10,10) | (B,C,6,10,10) | 1/16 |
6 | (B,320,3,5,5) | 320 |
表1 MSP-Net 配置信息
Tab. 1 Configuration information of MSP-Net
层序号 | 编码器输出特征图尺寸 | 通道数 | 解码器输出特征图尺寸 | 预测层输出特征图尺寸 | 损失函数权重 |
---|---|---|---|---|---|
1 | (B,32,96,160,160) | 32 | (B,32,96,160,160) | (B,C,96,160,160) | 1 |
2 | (B,64,48,80,80) | 64 | (B,64,48,80,80) | (B,C,48,80,80) | 1/2 |
3 | (B,128,24,40,40) | 128 | (B,128,24,40,40) | (B,C,24,40,40) | 1/4 |
4 | (B,256,12,20,20) | 256 | (B,256,12,20,20) | (B,C,12,20,20) | 1/8 |
5 | (B,320,6,10,10) | 320 | (B,320,6,10,10) | (B,C,6,10,10) | 1/16 |
6 | (B,320,3,5,5) | 320 |
网络结构 | DSC/% | IoU/% | HD95/mm | Precision/% | Recall/% | 首轮训练时长/s | 网络参数量 |
---|---|---|---|---|---|---|---|
Residual U-Net | 80.11 | 67.34 | 24.28 | 75.80 | 85.85 | 53.20 | 7 912 874 |
nnU-Net2D | 84.18 | 73.10 | 3.53 | 86.66 | 82.28 | 64.27 | 20 755 726 |
nnU-Net3D | 83.12 | 71.88 | 18.12 | 87.13 | 82.37 | 81.03 | 31 195 594 |
V-Net | 83.42 | 71.82 | 9.88 | 79.67 | 88.47 | 97.33 | 45 597 898 |
UNETR | 84.69 | 73.89 | 8.87 | 84.02 | 85.93 | 165.43 | 115 644 578 |
Swin UNETR2D | 82.55 | 70.84 | 3.60 | 84.51 | 81.50 | 168.43 | 25 137 980 |
Swin UNETR3D | 84.85 | 74.12 | 17.44 | 81.82 | 88.66 | 291.88 | 62 186 708 |
两阶段网络 | 84.46 | 73.40 | 13.32 | 80.82 | 89.14 | 304.76 | 72 707 046 |
本文网络 | 87.16 | 77.47 | 3.69 | 85.54 | 89.36 | 192.74 | 105 876 298 |
表2 不同网络预测结果的平均评价指标对比
Tab. 2 Comparison of average evaluation indexes of prediction results of different networks
网络结构 | DSC/% | IoU/% | HD95/mm | Precision/% | Recall/% | 首轮训练时长/s | 网络参数量 |
---|---|---|---|---|---|---|---|
Residual U-Net | 80.11 | 67.34 | 24.28 | 75.80 | 85.85 | 53.20 | 7 912 874 |
nnU-Net2D | 84.18 | 73.10 | 3.53 | 86.66 | 82.28 | 64.27 | 20 755 726 |
nnU-Net3D | 83.12 | 71.88 | 18.12 | 87.13 | 82.37 | 81.03 | 31 195 594 |
V-Net | 83.42 | 71.82 | 9.88 | 79.67 | 88.47 | 97.33 | 45 597 898 |
UNETR | 84.69 | 73.89 | 8.87 | 84.02 | 85.93 | 165.43 | 115 644 578 |
Swin UNETR2D | 82.55 | 70.84 | 3.60 | 84.51 | 81.50 | 168.43 | 25 137 980 |
Swin UNETR3D | 84.85 | 74.12 | 17.44 | 81.82 | 88.66 | 291.88 | 62 186 708 |
两阶段网络 | 84.46 | 73.40 | 13.32 | 80.82 | 89.14 | 304.76 | 72 707 046 |
本文网络 | 87.16 | 77.47 | 3.69 | 85.54 | 89.36 | 192.74 | 105 876 298 |
实验序号 | 网络配置 | DSC/ % | IoU/ % | HD95/ mm | Precision/ % | Recall/ % | ||
---|---|---|---|---|---|---|---|---|
MSE | MND | Deep | ||||||
1 | √ | √ | √ | 80.98 | 68.29 | 6.32 | 83.71 | 79.63 |
3 | √ | √ | 80.35 | 67.54 | 20.46 | 80.93 | 81.32 | |
5 | √ | 78.91 | 65.59 | 15.28 | 79.08 | 79.90 | ||
6 | √ | √ | 80.96 | 68.34 | 10.77 | 83.13 | 80.40 | |
7 | √ | √ | 80.96 | 68.37 | 18.14 | 82.23 | 81.18 |
表3 不同模块下实验预测结果平均评价指标的对比
Tab.3 Comparison of average evaluation indexes of prediction results under different modules
实验序号 | 网络配置 | DSC/ % | IoU/ % | HD95/ mm | Precision/ % | Recall/ % | ||
---|---|---|---|---|---|---|---|---|
MSE | MND | Deep | ||||||
1 | √ | √ | √ | 80.98 | 68.29 | 6.32 | 83.71 | 79.63 |
3 | √ | √ | 80.35 | 67.54 | 20.46 | 80.93 | 81.32 | |
5 | √ | 78.91 | 65.59 | 15.28 | 79.08 | 79.90 | ||
6 | √ | √ | 80.96 | 68.34 | 10.77 | 83.13 | 80.40 | |
7 | √ | √ | 80.96 | 68.37 | 18.14 | 82.23 | 81.18 |
实验序号 | 网络配置 | DSC/ % | IoU/ % | HD95/ mm | Precision/% | Recall/ % | ||
---|---|---|---|---|---|---|---|---|
MSEX2 | MNDT | MNDC | ||||||
1 | 80.98 | 68.29 | 6.32 | 83.71 | 79.63 | |||
2 | √ | 79.64 | 66.66 | 31.03 | 78.66 | 82.34 | ||
4 | √ | 80.25 | 67.35 | 21.17 | 79.85 | 82.22 | ||
8 | √ | 81.12 | 68.58 | 14.60 | 82.55 | 81.20 |
表4 不同配置下实验预测结果平均评价指标的对比
Tab.4 Comparison of average evaluation indexes of prediction results under different configurations
实验序号 | 网络配置 | DSC/ % | IoU/ % | HD95/ mm | Precision/% | Recall/ % | ||
---|---|---|---|---|---|---|---|---|
MSEX2 | MNDT | MNDC | ||||||
1 | 80.98 | 68.29 | 6.32 | 83.71 | 79.63 | |||
2 | √ | 79.64 | 66.66 | 31.03 | 78.66 | 82.34 | ||
4 | √ | 80.25 | 67.35 | 21.17 | 79.85 | 82.22 | ||
8 | √ | 81.12 | 68.58 | 14.60 | 82.55 | 81.20 |
1 | World Heart Federation. World heart report 2023: confronting the world’s number one killer [R/OL]. [2024-10-20].. |
2 | 《中国心血管健康与疾病报告2022》编写组.《中国心血管健康与疾病报告2022》概述[J]. 中国心血管病研究, 2023, 21(7): 577-600. |
The Writing Committee of the report on cardiovascular health and diseases in China 2022. Overview of report on cardiovascular health and diseases in China 2022[J]. Chinese Journal of Cardiovascular Research, 2023, 21(7): 577-600. | |
3 | KNUUTI J, WIJNS W, SARASTE A, et al. 2019 ESC Guidelines for the diagnosis and management of chronic coronary syndromes: the task force for the diagnosis and management of chronic coronary syndromes of the European Society of Cardiology (ESC)[J]. European Heart Journal, 2020, 41(3): 407-477. |
4 | GULATI M, LEVY P D, MUKHERJEE D, et al. 2021 AHA/ACC/ASE/CHEST/SAEM/SCCT/SCMR Guideline for the evaluation and diagnosis of chest pain: a report of the American college of cardiology/American heart association joint committee on clinical practice guidelines [J]. Circulation, 2021, 144(22): e368-e454. |
5 | 高扬,吕滨. 冠状动脉CT血管成像最新临床应用推荐及诊断规范[J].中华放射学杂志, 2022, 56(10): 1160-1164. |
GAO Y, LYU B. The latest clinical application recommendation and diagnostic criteria of coronary CT angiography[J]. Chinese Journal of Radiology, 2022, 56(10): 1160-1164. | |
6 | BENEDEK T, MESTER A, BENEDEK A, et al. Assessment of coronary plaque vulnerability in acute coronary syndromes using optical coherence tomography and intravascular ultrasound. a systematic review[J]. Journal of Cardiovascular Emergencies, 2016, 2(4): 173-184. |
7 | RODRIGUEZ-GRANILLO G A, CARRASCOSA P, BRUINING N, et al. Defining the non-vulnerable and vulnerable patients with computed tomography coronary angiography: evaluation of atherosclerotic plaque burden and composition[J]. European Heart Journal — Cardiovascular Imaging, 2016, 17(5): 481-491. |
8 | BARKANA B D, SARICICEK I, YILDIRIM B. Performance analysis of descriptive statistical features in retinal vessel segmentation via fuzzy logic, ANN, SVM, and classifier fusion [J]. Knowledge-Based Systems, 2017, 118: 165-176. |
9 | YANG H, ZHEN X, CHI Y, et al. CPR-GCN: conditional partial-residual graph convolutional network in automated anatomical labeling of coronary arteries[C]// Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 3802-3810. |
10 | 郭笑妍,梅雪,李振华,等. 形状约束下活动轮廓模型冠脉血管图像多尺度分割[J]. 中国图象图形学报, 2016, 21(7): 923-932. |
GUO X Y, MEI X, LI Z H, et al. Multi-scale segmentation of coronary artery image based on active contour model with shape constraints[J]. Journal of Image and Graphics, 2016, 21(7): 923-932. | |
11 | 罗宇杰,祝磊,马骏,等. 一种基于CTA图像的冠状动脉自动分割算法[J]. 传感技术学报, 2022, 35(11): 1491-1498. |
LUO Y J, ZHU L, MA J, et al. An automatic coronary artery segmentation algorithm based on CTA[J]. Chinese Journal of Sensors and Actuators, 2022, 35(11): 1491-1498. | |
12 | SHELHAMER E, LONG J, DARRELL T. Fully convolutional networks for semantic segmentation[C]// Proceedings of the 2015 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2015: 3431-3440. |
13 | RONNEBERGER O, FISCHER P, BROX T. U-Net: convolutional networks for biomedical image segmentation[C]// Proceedings of the 2015 International Conference on Medical Image Computing and Computer-Assisted Intervention, LNCS 9351. Cham: Springer, 2015: 234-241. |
14 | ÇIÇEK Ö, ABDULKADIR A, LIENKAMP S S, et al. 3D U-Net: learning dense volumetric segmentation from sparse annotation[C]// Proceedings of the 2016 International Conference on Medical Image Computing and Computer-Assisted Intervention, LNCS 9901. Cham: Springer, 2016: 424-432. |
15 | MILLETARI F, NAVAB N, AHMADI S A. V-Net: fully convolutional neural networks for volumetric medical image segmentation [C]// Proceedings of the 4th International Conference on 3D Vision. Piscataway: IEEE, 2016: 565-571. |
16 | MOU L, ZHAO Y, FU H, et al. CS2-Net: deep learning segmentation of curvilinear structures in medical imaging[J]. Medical Image Analysis, 2021, 67: No.101874. |
17 | WANG Y, WEI X, LIU F, et al. Deep distance transform for tubular structure segmentation in CT scans[C]// Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 3832-3841. |
18 | ZHOU Z, SIDDIQUEE M M R, TAJBAKHSH N, et al. UNet++: a nested U-Net architecture for medical image segmentation[C]// Proceedings of the 2018 International Workshop on Deep Learning in Medical Image Analysis, LNCS 11045. Cham: Springer, 2018: 3-11. |
19 | 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]. Nature Methods, 2021, 18(2): 203-211. |
20 | DOSOVITSKIY A, BEYER L, KOLESNIKOV A, et al. An image is worth 16×16 words: transformers for image recognition at scale[EB/OL]. [2024-06-22].. |
21 | HATAMIZADEH A, TANG Y, NATH V, et al. UNETR: Transformers for 3D medical image segmentation [C]// Proceedings of the 2022 IEEE/CVF Winter Conference on Applications of Computer Vision. Piscataway: IEEE, 2022: 1748-1758. |
22 | HATAMIZADEH A, NATH V, TANG Y, et al. Swin UNETR: Swin Transformers for semantic segmentation of brain tumors in MRI images[C]// Proceedings of the 2021 International MICCAI Brainlesion Workshop, LNCS 12962. Cham: Springer, 2022: 272-284. |
23 | JIANG W, LI Y, JIA Y, et al. Ori-Net: orientation-guided neural network for automated coronary arteries segmentation[J]. Expert Systems with Applications, 2024, 238(Pt C): No.121905. |
24 | CHEN F, LI S, WEI C, et al. Positive-unlabeled learning for coronary artery segmentation in CCTA images[J]. Biomedical Signal Processing and Control, 2024, 87(Pt B): No.105473. |
25 | NANNINI G, SAITTA S, BAGGIANO A, et al. A fully automated deep learning approach for coronary artery segmentation and comprehensive characterization[J]. APL Bioengineering, 2024, 8(1): No.016103. |
26 | GHARLEGHI R, ADIKARI D, ELLENBERGER K, et al. Automated segmentation of normal and diseased coronary arteries — the ASOCA challenge[J]. Computerized Medical Imaging and Graphics, 2022, 97: No.102049. |
27 | GHARLEGHI R, ADIKARI D, ELLENBERGER K, et al. Annotated computed tomography coronary angiogram images and associated data of normal and diseased arteries[J]. Scientific Data, 2023, 10: No.128. |
28 | MEDRANO-GRACIA P, ORMISTON J, WEBSTER M, et al. A computational atlas of normal coronary artery anatomy[J]. EuroIntervention, 2016, 12(7): 845-854. |
29 | MEDRANO-GRACIA P, ORMISTON J, WEBSTER M, et al. Construction of a coronary artery atlas from CT angiography[C]// Proceedings of the 2014 International Conference on Medical Image Computing and Computer-Assisted Intervention, LNCS 8674. Cham: Springer, 2014: 513-520. |
30 | ULYANOV D, VEDALDI A, LEMPITSKY V. Instance normalization: the missing ingredient for fast stylization[EB/OL]. [2024-10-22].. |
31 | MAAS A L, HANNUN A Y, NG A Y. Rectifier nonlinearities improve neural network acoustic models[EB/OL]. [2024-10-22].. |
32 | LEE C Y, XIE S, GALLAGHER P W, et al. Deeply-supervised nets[C]// Proceedings of the 18th International Conference on Artificial Intelligence and Statistics. New York: JMLR.org, 2015: 562-570. |
33 | SZEGEDY C, LIU W, JIA Y, et al. Going deeper with convolutions[C]// Proceedings of the 2015 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2015: 1-9. |
34 | SZEGEDY C, IOFFE S, VANHOUCKE V, et al. Inception-v4, Inception-ResNet and the impact of residual connections on learning[C]// Proceedings of the 31st AAAI Conference on Artificial Intelligence. Palo Alto: AAAI Press, 2017: 4278-4284. |
35 | KERFOOT E, CLOUGH J, OKSUZ I, et al. Left-ventricle quantification using residual U-Net[C]// Proceedings of the 2018 International Workshop on Statistical Atlases and Computational Models of the Heart, LNCS 11395. Cham: Springer, 2019: 371-380. |
36 | YUSHKEVICH P A, PIVEN J, HAZLETT H C, et al. User-guided 3D active contour segmentation of anatomical structures: significantly improved efficiency and reliability[J]. NeuroImage, 2006, 31(3): 1116-1128. |
[1] | 周阳, 李辉. 基于语义和细节特征双促进的遥感影像建筑物提取网络[J]. 《计算机应用》唯一官方网站, 2025, 45(4): 1310-1316. |
[2] | 李晨倩, 刘俊. 基于半监督和多尺度级联注意力的超声颈动脉斑块分割方法[J]. 《计算机应用》唯一官方网站, 2024, 44(8): 2604-2610. |
[3] | 王美, 苏雪松, 刘佳, 殷若南, 黄珊. 时频域多尺度交叉注意力融合的时间序列分类方法[J]. 《计算机应用》唯一官方网站, 2024, 44(6): 1842-1847. |
[4] | 张鹏飞, 韩李涛, 冯恒健, 李洪梅. 基于注意力机制和全局特征优化的点云语义分割[J]. 《计算机应用》唯一官方网站, 2024, 44(4): 1086-1092. |
[5] | 王铂越, 李英祥, 钟剑丹. 基于改进Res-UNet的昼夜地基云图分割网络[J]. 《计算机应用》唯一官方网站, 2024, 44(4): 1310-1316. |
[6] | 付顺旺, 陈茜, 李智, 王国美, 卢妤. 用于篡改图像检测和定位的双通道渐进式特征过滤网络[J]. 《计算机应用》唯一官方网站, 2024, 44(4): 1303-1309. |
[7] | 李威, 陈玲, 徐修远, 朱敏, 郭际香, 周凯, 牛颢, 张煜宸, 易珊烨, 章毅, 罗凤鸣. 基于多任务学习的间质性肺病分割算法[J]. 《计算机应用》唯一官方网站, 2024, 44(4): 1285-1293. |
[8] | 吴宁, 罗杨洋, 许华杰. 基于多尺度特征融合的遥感图像语义分割方法[J]. 《计算机应用》唯一官方网站, 2024, 44(3): 737-744. |
[9] | 李子怡, 曲婷婷, 崇乾鹏, 徐金东. 基于模糊多尺度特征的遥感图像分割网络[J]. 《计算机应用》唯一官方网站, 2024, 44(11): 3581-3586. |
[10] | 刘永江, 陈斌. 基于多尺度记忆库的像素级无监督工业异常检测[J]. 《计算机应用》唯一官方网站, 2024, 44(11): 3587-3594. |
[11] | 郑秋梅, 牛薇薇, 王风华, 赵丹. 基于细节增强的双分支实时语义分割网络[J]. 《计算机应用》唯一官方网站, 2024, 44(10): 3058-3066. |
[12] | 周迪, 张自力, 陈佳, 胡新荣, 何儒汉, 张俊. 基于EfficientNetV2和物体上下文表示的胃癌图像分割方法[J]. 《计算机应用》唯一官方网站, 2023, 43(9): 2955-2962. |
[13] | 齐爱玲, 王宣淋. 基于中层细微特征提取与多尺度特征融合细粒度图像识别[J]. 《计算机应用》唯一官方网站, 2023, 43(8): 2556-2563. |
[14] | 郑帅, 张晓龙, 邓鹤, 任宏伟. 基于多尺度特征融合和网格注意力机制的三维肝脏影像分割方法[J]. 《计算机应用》唯一官方网站, 2023, 43(7): 2303-2310. |
[15] | 鲁斌, 柳杰林. 基于特征增强的三维点云语义分割[J]. 《计算机应用》唯一官方网站, 2023, 43(6): 1818-1825. |
阅读次数 | ||||||
全文 |
|
|||||
摘要 |
|
|||||