基于双流神经网络的颈动脉粥样硬化斑块稳定性区分方法

doi:10.3969/j.issn.1673-5765.2019.05.007

摘要/Abstract

摘要：

目的训练双流神经网络自动区分颈动脉粥样硬化斑块的稳定性。方法使用颈动脉内膜剥脱术后经病理证实的115例稳定颈动脉粥样硬化斑块患者和110例易损颈动脉粥样硬化斑块患者的844段超声造影视频，将其中744段视频作为训练集，训练双流神经网络，得到在训练集上区分效果最佳的神经网络。将剩余的100段视频作为测试集，测试该神经网络自动区分颈动脉粥样硬化斑块稳定性的价值。结果神经网络在训练集上区分颈动脉斑块稳定性的准确率、敏感度、特异度、阳性预测值、阴性预测值、阳性似然比、阴性似然比分别为93%、87%、97%、96%、90%、29和0.13，在测试集上相应的结果分别为80%、70%、90%、88%、75%、7和0.33。受试者工作特征曲线上，训练集和测试集中双流神经网络判断斑块易损性的曲线下面积分别为0.99和0.84，均P＜0.001。结论利用已知病理结果的超声造影视频，将其输入到双流神经网络进行训练，能得到较好的自动区分颈动脉粥样硬化斑块稳定性的模型。

文章导读： 本研究以病理诊断结果为金标准，训练双流神经网络学习颈动脉粥样硬化斑块的超声造影视频特点，以期为临床更高效、精准的判断超声影像中显示的动脉粥样硬化斑块的性质提供新的思路和技术。

关键词: 颈动脉; 斑块; 超声造影; 双流神经网络

Abstract:

Objective To train two-stream neural network to distinguish the stability of carotid plaques. Methods 844 contrast-enhanced ultrasound videos were used in the experiment. They were from 115 patients with stable carotid plaques and 110 patients with vulnerable carotid plaques verified by pathology after CEA. 744 videos were used as training set to train two-stream network, to find the neural network segment having optimal recognition effect. The left 100 videos were used as test set to distinguish the stability of carotid plaque. Results Accuracy, sensitivity, specificity, positive predictive value, negative predictive value, positive likelihood ratio and negative likelihood ratio in training set were 93%, 87%, 97%, 96%, 90%, 29 and 0.13, respectively. The corresponding results in test set were 80%, 70%, 90%, 88%, 75%, 7 and 0.33, respectively. Area under the receiver operating characteristic curve for training set and test set were 0.99 and 0.84 (both P <0.001). Conclusions Training two-stream neural network with contrast-enhanced ultrasound videos with known pathological results of plaques can obtain a model of recognizing the stability of carotid plaques.

Key words: Carotid artery; Plaque; Ultrasonic radiography; Two-stream neural network

宁彬，李璐，于腾飞，童挥，何文，赵明昌. 基于双流神经网络的颈动脉粥样硬化斑块稳定性区分方法[J]. 中国卒中杂志, 2019, 14(05): 438-443.

NING Bin, LI Lu,YU Teng-Fei, TONG Hui, HE Wen, ZHAO Ming-Chang. Method of Distinguishing Stability of Carotid Plaque Based on Two-stream Neural Network[J]. Chinese Journal of Stroke, 2019, 14(05): 438-443.

参考文献

[1] 宁彬，何文，张东，等. 斑块内新生血管分布特征的

超声造影与病理对照[J]. 中国医学影像技术，2015，

31（5）：655-658.

[2] KANBER B，HARTSHORNE T C，HORSFIELD

M A，et al. Quantitative assessment of carotid

plaque surface irregularities and correlation to

cerebrovascular symptoms[J/OL]. Cardiovasc

Ultrasound，2013，11：38. https：//doi.

org/10.1186/1476-7120-11-38.

[3] DOU Q，CHEN H，YU L，et al. Automatic detection

of cerebral microbleeds from MR images via 3D

convolutional neural networks[J]. IEEE Trans Med

Imaging，2016，35（5）：1182-1195.

[4] CHEN H，NI D，QIN J，et al. Standard plane

localization in fetal ultrasound via domain

transferred deep neural networks[J]. IEEE J Biomed

Health Inform，2015，19（5）：1627-1636.

[5] SHIN S Y，LEE S，YUN I D，et al. Joint weakly

and Semi-Supervised deep learning for localization

and classification of masses in breast ultrasound

images[J/OL]. IEEE Trans Med Imaging，2018，12.

https：//doi.org/10.1109/TMI.2018.2872031.

[6] LEKADIR K，GALIMZIANOVA A，BETRIU A，

et al. A convolutional neural network for automatic

characterization of plaque composition in carotid

ultrasound[J]. IEEE J Biomed Health Inform，2016，

21（1）：48.

[7] 孙夏，吴蔚，吴鹏，等. 基于卷积神经网络的颈动脉

斑块超声图像特征识别[J]. 中国医疗器械信息，2016，

22（9）：4-8.

[8] 赵媛，孙夏，丁明跃. 一种基于深度学习的颈动脉斑

块超声图像识别方法[J]. 中国医疗器械信息，2017，

23（9）：9-11.

[9] SIMONYAN K，ZISSERMAN A. Two-Stream

convolutional networks for action recognition in

videos[C]. Montreal，28th Conference on Neural

Information Processing Systems NIPS，2014.

[10] GUNNAR F. Two-Frame motion estimation based

on polynomial expansion[C]. Halmstad，Sweden，

13th Scandinavian Conference on Image Analysis

（SCIA 2003），2003.

[11] YOSINSKI J，CLUNE J，BENGIO Y，et al. How transferable are features in deep neural networks?[J].

Eprint Arxiv，2014，27（11）：3320-3328.

[12] FEINSTEIN S B. Contrast ultrasound imaging of

the carotid artery vasa vasorum and atherosclerotic

plaque neovascularization[J]. J Am Coll Cardiol，

2006，48（2）：236-243.

[13] VIRMANI R，KOLODGIE F D，BURKE A P，et al.

Atherosclerotic plaque progression and vulnerability

to rupture：angiogenesis as a source of intraplaque

hemorrhage[J]. Arterioscler Thromb Vasc Biol，2005，

25（10）：2054-2061.

[14] COLI S，MAGNONI M，SANGIORGI G，et

al. Contrast-Enhanced ultrasound imaging of

intraplaque neovascularization in carotid arteries[J].

J Am Coll Cardiol，2008，52（3）：223-230.

[15] STAUB D，SCHINKEL A F，COLL B，et al.

Contrast-Enhanced Ultrasound Imaging of the

Vasa Vasorum：from early atherosclerosis to

the identification of unstable plaques[J]. JACC：

Cardiovasc Imag，2010，3（7）：761-771.