基于深度学习的人工智能在肺结节诊断领域的进展

张俊; 侯聪; 刘新疆

doi:10.12122/j.issn.1674-4500.2020.03.01

基于深度学习的人工智能在肺结节诊断领域的进展

doi: 10.12122/j.issn.1674-4500.2020.03.01

张俊^1,,
侯聪¹,
刘新疆^1,2, ,

1.
滨州医学院附属医院放射科，山东滨州 256603
2.
上海市浦东医院（复旦大学附属浦东医院）放射科，上海

基金项目:

山东省自然科学基金计划 ZR2016HL43

山东省医药卫生科技发展计划 2015WS0479

详细信息

作者简介:
张俊，在读硕士研究生，住院医师，E-mail: zj20120521@126.com

通讯作者:
刘新疆，博士，主任医师，E-mail: lxj6513@163.com

计量
- 文章访问数: 1488
- HTML全文浏览量: 544
- PDF下载量: 65
- 被引次数: 0
出版历程
- 收稿日期: 2020-07-06
- 刊出日期: 2020-07-15

Research progress of artificial intelligence based on deep learning in the field of pulmonary nodules

Jun ZHANG^1
,,
Cong HOU¹,
Xinjiang LIU^{1,2
, ,}

1.
Department of Radiology, Affiliated Hospital of Binzhou Medical University, Binzhou 256603, China
2.
Department of Radiology, Shanghai Pudong Hospital (Pudong Hospital Affiliated to Fudan University), Shanghai 201399, China

摘要

摘要: 肺癌是死亡率最高的恶性肿瘤，肺结节的早期检测是降低肺癌死亡率的关键。基于深度学习的人工智能技术可通过自我学习，不断提高肺结节检测和诊断的准确率，是实现计算机辅助诊断的重要手段。本文介绍了人工智能、机器学习、深度学习的概念及三者间的关系，阐述了4种常见的深度学习模型：卷积神经网络、海量训练人工神经网络、自编码器和深度信念网络。卷积神经网络是最常用的深度学习模型，主要包括二维卷积神经网络、三维卷积神经网络和多流、多尺度的卷积神经网络，其中的多流、多尺度的卷积神经网络更有利于肺结节的分类；海量训练人工神经网络在有限的肺结节训练样本中具有优势；自编码器可以在较低维空间下对肺结节进行检测；深度信念网络是一种生成模式，与极限学习机结合可提高肺结节的诊断率。另外，本研究分析了目前人工智能存在的问题：标记图像过少、可解释性和可控制性不足、存在伦理和法律问题。总之，基于深度学习的人工智能不仅改变了影像学，也改变了所有其他的医学领域，具有广阔的应用前景。
- 深度学习 /
- 机器学习 /
- 人工智能 /
- 肺结节 /
- 神经网络
Abstract: Lung cancer is the malignant tumor with the highest mortality rate. Early diagnosis of lung nodules is the key to reducing the mortality of lung cancer. Artificial intelligence technology based on deep learning can continuously improve the accuracy of lung nodule detection and diagnosis through self-learning, which is an important means to achieve computer-aided diagnosis. This article first briefly introduces the concepts of artificial intelligence, machine learning, deep learning, and the relationship between the three. The paper describes four common deep learning models: convolutional neural network, massive-training artificial neural network, auto-encoder, and deep belief network. The convolutional neural network is the most commonly used deep learning model, mainly including two dimensional convolutional neural network, three dimensional convolutional neural network and multi-stream multi-scale convolutional neural network, of which multi-stream multi-scale convolutional neural network it is more conducive to the classification of lung nodules. The massive-training artificial neural network has advantages in limited lung nodule training samples. The auto-encoder can detect lung nodules in a lower dimensional space.The deep belief network is a generation mode. Combining with extreme learning machine could improve the diagnosis rate of pulmonary nodules. Finally, it analyzes the current problems of artificial intelligence: too few labeled images; insufficient interpretability and controllability. There are ethical and legal issues. In short, artificial intelligence based on deep learning has changed not only imaging, but also all other medical fields, and has broad application prospects.
- deep learning /
- machine learning /
- artificial intelligence /
- pulmonary nodules /
- neural network

HTML全文

参考文献(23)

[1]	Ferlay J, Soerjomataram I, Dikshit R, et al. Cancer incidence and motality worldwide: sources, methods and major patterns in GLOBOCAN 2012[J]. Int J Cancer, 2015, 136(5): E359-86. doi: 10.1002/ijc.29210
[2]	李欣菱, 郭芳芳, 周振, 等.基于深度学习的人工智能胸部CT肺结节检测效能评估[J].中国肺癌杂志, 2019, 22(6): 336-40. http://d.old.wanfangdata.com.cn/Periodical/zgfazz201906002
[3]	Hua KL, Hsu CH, Hidayati SC, et al. Computer-aided classificationof lung nodules on computed tomography images via deep learning technique[J]. Onco Targets Ther, 2015, 8: 2015-22. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4531007/
[4]	Liu S, Xie YT, Jirapatnakul A, et al. Pulmonary nodule classification in lung cancer screening with three-dimensional convolutional neural networks [J]. J Med Imaging (Bellingham), 2017, 4(4): 308-16. http://www.ncbi.nlm.nih.gov/pubmed/29181428
[5]	Ciompi F, Chung K, van Riel SJ, et al. Corrigendum: Towards automatic pulmonary nodule management in lung cancer screening with deep learning[J]. Sci Rep, 2017, 7: 46878-85. doi: 10.1038/srep46878
[6]	Shen W, Zhou M, Yang F, et al. Multi-scale convolutional neural networks for lung nodule classification[J]. Inf Process Med Imaging, 2015, 24: 588-99. doi: 10.1007%2F978-3-319-19992-4_46
[7]	Suzuki K, DoiK. How can a massive training artificial neural network (MTANN) be trained with a small number of cases in the distinction between nodules and vessels in thoracic CT[J]. Acad ·Radiol, 2005, 12(10): 1333-41. doi: 10.1016/j.acra.2005.06.017
[8]	Nima T, Kenji S. Comparing two classes of end-to-end machinelearning models in lung nodule detection and classification: MTANNs vs. CNNs [J]. Pattern Recogn, 2017, 63: 476-86. doi: 10.1016/j.patcog.2016.09.029
[9]	Luca S, Mainak B, Venkatanareshbabu K, et al. The present and future of deep learning in radiology[J]. Eur J Radiol, 2019, 114: 14-24. doi: 10.1016/j.ejrad.2019.02.038
[10]	Bengio Y, Lamblin P, Popovici D, et al. Greedy layer-wise training of deep networks[C]. Canada: Proceedings of the 19th International Conference on Neural Information Processing System, 2006: 153-60.
[11]	Vincent P, Larochelle H, Bengio Y, et al. Extracting and composing robust features with denoising autoencoders [C]. New York: Proceedings of the 25th International Conference on Machine Learning, 2008: 1096-103.
[12]	Hinton GE, Osindero S, Teh YW. A fast learning algorithm for deep belief nets [J]. Neural Comput, 2006, 18(7): 1527-54. doi: 10.1162/neco.2006.18.7.1527
[13]	Kumar D, Wong A, Clausi DA. Lung nodule classification using deep features in CT images[C]//2015 12th Conference on Computer and Robot Vision, 3-5 June 2015, Halifax, NS, Canada. IEEE, 2015: 133-8.
[14]	Mao KM, Tang RJ, Wang XQ, et al. Feature representation using deep autoencoder for lung nodule image classification[J]. Complexity, 2018, 128 (13): 1-11. doi: 10.1155/2018/3078374
[15]	赵鑫, 强彦, 强梓林, 等.基于局部感受野和半监督深度自编码的肺结节检测方法[J].科学技术与工程, 2017, 17(33): 125-30. doi: 10.3969/j.issn.1671-1815.2017.33.018
[16]	Hinton GE, Salakhutdinov RR. Reducing the dimensionality of data with neural networks [J]. Science, 2006, 313(5786): 504-7. doi: 10.1126-science.1127647/
[17]	Litjens G, Kooi T, Bejnordi BE, et al. A survey on deep learning in medical image analysis [J]. Med Image Anal, 2017, 42: 60-88. doi: 10.1016/j.media.2017.07.005
[18]	高琰, 陈白帆, 晁绪耀, 等.基于对比散度-受限玻尔兹曼机深度学习的产品评论情感分析[J].计算机应用, 2016, 36(4): 1045-9. http://d.old.wanfangdata.com.cn/Periodical/jsjyy201604031
[19]	张婷, 赵文婷, 赵涓涓, 等.改进的深度信念网络肺结节良恶性分类[J].计算机工程与设计, 2018, 39(9): 2707-13, 2729. http://d.old.wanfangdata.com.cn/Periodical/jsjgcysj201809002
[20]	Hussein S, Cao K, Song Q, et al. Risk stratification of lung nodules using 3d CNN-based multi-task learning[C]//International Conference on Information Processing in Medical Imaging. Boone: IPMI, 2017: 249-60.
[21]	Zamir AR, Sax A, Shen W, et al. Taskonomy: Disentangling task transfer learning[C]// 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE, 2018: 3712-22.
[22]	Zhang C, Sun X, Dang K, et al. Toward an expert level of lung cancer detection and classification using a deep convolutional neural network[J]. Oncologist, 2019, 24(9): 1159-65. https://pubmed.ncbi.nlm.nih.gov/30996009/
[23]	Sala E, Mema E, Himoto Y, et al. Unravelling tumour heterogeneity using next-generation imaging: radiomics, radiogenomics, and habitat imaging[J]. Clin Radiol, 2017, 72(1): 3-10. https://www.ncbi.nlm.nih.gov/pubmed/27742105