一种基于 OCT 图像的深度学习 I 和 II 型 wAMD辅助分型方法

doi:10.3760/cma.j.cn115909-20210204-00046

摘要
图/表
参考文献(15)
相关文章 (15)

全文: PDF 2316KB HTML (1 KB)
输出: BibTeX | EndNote (RIS) 背景资料

文章导读

摘要目的：探讨基于光学相干断层扫描（OCT）图像的深度学习算法用于湿性年龄相关性黄斑变性（wAMD）的分型的可行性以达到辅助眼科疾病诊断的目的。方法：数据使用2018年6月至2019年 6月在宁波市眼科医院门诊确诊为wAMD患者39例（46眼）。首先将资深眼科专家提供的每例患者 AMD等级作为分型算法的金标准，再使用Resnet34模型结构输出预测分型结果并与金标准对比，不断进行参数微调直到损失收敛，最终实现自动判断患者疾病等级。结果：使用已训练的模型进行所有测试病例的预测，基于正确的病例数与所有测试病例数的比例作为算法最终的准确率，结果表明深度学习网络方法分型准确率，与普通医师wAMD分型准确率相比，高出20%。通过梯度加权的类激活映射可视化模型诊断依据可作为临床医师诊断的参考。结论：深度学习经OCT图像数据训练后对wAMD分型的准确率明显高于普通医师对wAMD的分型准确率。基于深度学习算法的wAMD分型结果可用于疾病的辅助诊断，缓解国内专业眼科医师紧缺的现状。

	服务

	把本文推荐给朋友
	加入我的书架
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章

关键词 ：深度学习, 湿性年龄相关性黄斑变性, 病灶分型, 光学相干断层扫描

Abstract： Objective: To investigate the application of a deep learning algorithm based on optical coherence tomography (OCT) images for the classification of wet age-related macular degeneration (wAMD) so as to assist in the diagnosis of ophthalmic diseases. Methods: Data of this study were collected from 39 patients (46 eyes) diagnosed with wAMD in the outpatient department of Ningbo Eye Hospital from June 2018 to June 2019. First, senior ophthalmologists provide the wAMD grade of each patient as the gold standard of the classification algorithm. Then, Resnet 34 was used to output the predicted results of the type of wAMD in order to compare these results with the gold standard. The parameters were fine-tuned continuously until the status of the loss was in convergence. Finally, the wAMD grade for patients was diagnosed automatically. Results: The accuracy of the algorithm is the proportion of the cases correctly predicted in all test cases. The experimental results showed that the accuracy rate of identifying the type of wAMD based on the deep learning network was 20% higher than for general ophthalmologists. Gradientweighted class activation mapping can visualize how the model can be used as a diagnostic reference for ophthalmologists. Conclusion: In the aspects of classifying wAMD by an adequately trained deep learning algorithm, its accuracy is significantly better than that of general ophthalmologists. Therefore, the classification results based on deep learning algorithms can be used for auxiliary diagnosis of diseases, so as to alleviate the scarcity of domestic ophthalmologists.

Key words： deep learning wet age-related macular degeneration lesion classification optical coherence tomography

收稿日期: 2021-02-04

PACS:

基金资助: 宁波市自然基金项目（2019A610354）；浙江省自然科学基金（LQ19H180001）；浙江省自然基金一般项目（LY19H120001）；宁波市公益计划面上项目（2019C50049）；2021年度浙江省中医药科技计划（2021ZB268）

通讯作者: 胡衍（ORCID：0000-0002-7663-3096），Email：huy3@sustech.edu.cn

引用本文:

沈俊勇1 龚雁2 胡衍1 廖燕红2 杨建龙3 赵一天3 刘江1，3. 一种基于 OCT 图像的深度学习 I 和 II 型 wAMD辅助分型方法[J]. 中华眼视光学与视觉科学杂志, 2021, 23(8): 615-621. Junyong Shen1,Yan Gong2,Yan Hu1,Yanhong Liao2,Jianlong Yang3,Yitian Zhao3,Jiang Liu1, 3. I and II wAMD-Aided Deep-Learning Grading Algorithm Based on OCT. Chinese Journal of Optometry Ophthalmology and Visual science, 2021, 23(8): 615-621. DOI: 10.3760/cma.j.cn115909-20210204-00046

链接本文:

https://www.cjoovs.com/CN/10.3760/cma.j.cn115909-20210204-00046 或 https://www.cjoovs.com/CN/Y2021/V23/I8/615

[1]	Bressler NM, Bressler SB, Fine SL. Age-related macular degeneration. Surv Ophthalmol, 1988, 32(6): 375-413. DOI: 10.1056/NEJMoa054481.
[2]	Lim LS, Mitchell P, Seddon JM, et al. Age-related macular degeneration. The Lancet, 2012, 379(9827): 1728-1738. DOI: 10.1016/S0140-6736(12)60282-7.
[3]	de Jong EK, Geerlings MJ, den Hollander AI. Age-related macular degeneration, genetics and genomics of eye disease, 2020: 155-180. DOI: 10.1016/B978-0-12-816222-4.00010-1.
[4]	Bird AC, Bressler NM, Bressler SB, et al. An international classification and grading system for age-related maculopathy and age-related macular degeneration. Surv Ophthalmol, 1995, 39(5): 367-374. DOI: 10.1016/S0039-6257(05)80092-X.
[5]	王艺晓, 姚腾腾, 童尧, 等. AAV介导的EPO基因下调对CNV 的抑制作用. 中华眼视光学与视觉科学杂志, 2019, 21(9): 641-647. DOI: 10.3760/cma.j.issn.1674-845X.2019.09.001.
[6]	Hee MR, Baumal CR, Puliafito CA, et al. Optical coherence tomography of age-related macular degeneration and choroidal neovascularization. Ophthalmology, 1996, 103(8): 1260-1270. DOI: 10.1016/S0161-6420(96)30512-5.
[7]	Coscas F, Puche N, Coscas G, et al. Comparison of macular choroidal thickness in adult onset foveomacular vitelliform dystrophy and age-related macular degeneration. Invest Ophthalmol Vis Sci, 2014, 55(1): 64-69. DOI: 10.1167/ iovs.13-12931.
[8]	中华医学会眼科学分会眼底病学组中国老年性黄斑变性临 . 中国老年性黄斑变性临床诊断治疗路径 . 中华眼底病杂志, 2013, 29(4): 343-355. DOI: 10.3760/cma. j.issn.1005-1015.2013.04.002.
[9]	He K, Zhang X, Ren S, et al. Deep residual learning for image recognition. CVPR, 2016: 770-778. DOI: 10.1109/ CVPR.2016.90.
[10]	沈梦溪, 汪枫桦, 孙晓东. 基于电子化信息平台的新生血管性年龄相关性黄斑变性真实世界研究进展. 中华眼视光学与视觉科学杂志, 2019, 21(3): 237-240. DOI: 10.3760/cma. j.issn.1674-845X.2019.03.015.
[11]	Lee CS, Baughman DM, Lee AY. Deep learning is effective for the classification of OCT images of normal versus age-related macular degeneration. Ophthalmol Retina, 2017, 1(4): 322-327. DOI: 10.1016/j.oret.2016.12.009.
[12]	Grassmann F, Mengelkamp J, Brandl C, et al. A deep learning algorithm for prediction of age-related eye disease study severity scale for age-related macular degeneration from color fundus photography. Ophthalmology, 2018, 125(9): 1410-1420. DOI: 10.1016/j.ophtha.2018.02.037.
[13]	Treder M, Lauermann JL, Eter N. Automated detection of exudative age-related macular degeneration in spectral domain optical coherence tomography using deep learning. Graefes Arch Clin Expe Ophthalmol, 2018, 256(2): 259-265. DOI: 10.1007/ s00417-017-3850-3.
[14]	龚雁, 顾在旺, 胡衍, 等．基于深度学习OCT辅助诊断湿性年龄相关性黄斑变性算法的应用. 中华实验眼科杂志, 2019, 37(8): 658-662. DOI: 10.3760/cma.j.issn.2095-0160.2019.08. 014.
[15]	Ferris FL, Fine SL, Hyman L. Age-related macular degeneration and blindness due to neovascularmaculopathy. Arch Ophthalmol, 1984, 102(11): 1640-1642. DOI: 10.1001/ archopht.1984.01040031330019.