Please wait a minute...
浙江大学学报(工学版)
浙江大学学报(工学版)  2023, Vol. 57 Issue (12): 2448-2455    DOI: 10.3785/j.issn.1008-973X.2023.12.012
计算机技术     
眼底病变OCT图像的轻量化识别算法
侯小虎1,2(),贾晓芬1,3,*(),赵佰亭2
1. 安徽理工大学第一附属医院(淮南市第一人民医院)，安徽 淮南 232001
2. 安徽理工大学 电气与信息工程学院，安徽 淮南 232001
3. 安徽理工大学 人工智能学院，安徽 淮南 232001
Lightweight recognition algorithm for OCT images of fundus lesions
Xiao-hu HOU1,2(),Xiao-fen JIA1,3,*(),Bai-ting ZHAO2
1. The First Affiliated Hospital of Anhui University of Science and Technology (Huainan First People's Hospital), Huainan 232001, China
2. Institute of Electrical and Information Engineering, Anhui University of Science and Technology, Huainan 232001, China
3. Institute of Artificial Intelligence, Anhui University of Science and Technology, Huainan 232001, China
 全文: PDF(1143 KB)   HTML
摘要：

为了准确、方便地识别多类型眼底病变，提出光学相干断层扫描技术（OCT）图像的轻量化分类模型MB-CNN. 降低卷积核的使用个数，调节每个阶段卷积块的使用比例，设计轻量化主干网络L-Resnet，通过加深网络深度增强对深层语义信息的提取. 使用深度可分离卷积设计多尺度卷积块MultiBlock，利用MultiBloc深度挖掘病灶区域的特征，使用不同的卷积核提取不同尺寸病变的特征，提高网络对病变OCT图像的识别能力. 构建特征融合模块FFM，融合浅层信息和深层信息，充分提取病变特征的纹理和语义信息，提高对小目标病变的识别能力. 实验结果显示，MB-CNN在UCSD、Duke和NEH3个数据集上的总体分类精度分别达到97.2%、99.92%和94.37%，模型参数量明显降低，所提模型能够针对眼底的多种病变进行分类.
关键词： 眼底病变;  光学相干断层扫描技术(OCT)图像;  智能识别;  轻量化分类模型;  语义信息;  特征融合    
Abstract:

A lightweight classification model MB-CNN for optical coherence tomography (OCT) images was proposed to accurately and conveniently identify multiple types of fundus lesions. By reducing the number of convolution cores and adjusting the proportion of convolution blocks in each stage, a lightweight backbone network L-Resnet was designed, and the extraction of deep-layer semantic information was enhanced by deepening the network depth. The multi-scale convolution block MultiBlock was designed using depthwise seperable convolution, and the features of the lesion area was mined. Different convolution kernels were used to extract the lesions features of different sizes to improve the recognition ability of the network to the OCT image of the lesion. The feature fusion module FFM was constructed, and the shallow layer information and deep layer information were fused, the texture and semantic information of the pathological features were extracted, and the recognition ability of small target lesions was improved. Experimental result showed that the overall classification accuracy of MB-CNN in the three datasets of UCSD, Duke and NEH was 97.2%, 99.92% and 94.37% respectively, the amount of model parameters were significantly reduced. The proposed model can classify various fundus lesions.
Key words: fundus lesions    optical coherence tomography (OCT) images    intelligent identification    lightweigh classification model    semantic information    feature fusion
收稿日期: 2023-03-23 出版日期: 2023-12-27
CLC:  TP 391.41  
基金资助: 安徽理工大学医学专项培育项目(YZ2023H2B006)；安徽理工大学引进人才科研启动基金资助项目(2022yjrc44)；国家自然科学基金资助项目(52174141)；安徽省自然科学基金资助项目(2108085ME158)；安徽理工大学研究生创新基金资助项目(2022CX2086)
通讯作者: 贾晓芬     E-mail: hxh19855424153@163.com;jxfzbt2008@163.com
作者简介: 侯小虎（1998—），男，硕士生，从事医学图像处理研究. orcid.org/0009-0005-8154-6578. E-mail： hxh19855424153@163.com
服务　　
把本文推荐给朋友
加入引用管理器
E-mail Alert
作者相关文章　　
侯小虎
贾晓芬
赵佰亭

引用本文:

侯小虎,贾晓芬,赵佰亭. 眼底病变OCT图像的轻量化识别算法[J]. 浙江大学学报(工学版), 2023, 57(12): 2448-2455.

Xiao-hu HOU,Xiao-fen JIA,Bai-ting ZHAO. Lightweight recognition algorithm for OCT images of fundus lesions. Journal of ZheJiang University (Engineering Science), 2023, 57(12): 2448-2455.

链接本文:

https://www.zjujournals.com/eng/CN/10.3785/j.issn.1008-973X.2023.12.012    或    https://www.zjujournals.com/eng/CN/Y2023/V57/I12/2448

图 1  轻量化分类模型MB-CNN的整体结构图
图 2  MultiBlock结构图
图 3  2种模块的结构图
图 4  OCT例图
实验 L-Resnet DW Conv MultiBlock FFM NP/106 Acc/%
1 — — — — 11.180 94.41
2 √ — — — 6.840 94.24
3 — √ — — 0.312 93.03
4 — — √ — 0.344 93.47
5 √ — √ — 0.521 94.11
6 √ — √ √ 1.350 94.57
表 1  MB-CNN组成部分的消融实验
模型 图片类别 P/% R/% Spe/% Acc/% NP/106 OAcc/%
LACNN[5] CNV 93.5 89.8 95.1 92.7 — 90.1±1.2
DME 86.4 87.5 98.0 96.6
玻璃膜疣 70.0 72.5 95.9 93.6
正常眼底 94.8 97.3 97.4 97.4
Multi-Label CNN[6] CNV 93.5 88.1 96.0 93.8 — 90.4±1.2
DME 83.1 86.0 95.8 96.5
玻璃膜疣 69.8 72.1 96.8 93.3
正常眼底 95.1 96.1 97.9 96.8
PCAM[7] — 91.52 91.22 — — 11.09 94.15±1.15
Resnet18[12] CNV 97.83 97.63 98.27 97.97 11.18 96.52±0.5
DME 95.30 95.17 99.27 98.70
玻璃膜疣 89.87 90.70 98.80 97.97
正常眼底 97.43 97.43 98.83 98.4
DMF-CNN[8] CNV 97.05 97.33 — — — 94.37
DME 96.26 93.22 94.64
玻璃膜疣 87.73 98.29 94.43
正常眼底 97.49 97.62 96.03
MB-CNN CNV 98.39 98.49 98.60 98.51 1.35 97.22±0.6
DME 97.30 96.27 99.49 99.07
玻璃膜疣 92.24 90.84 99.12 98.33
正常眼底 97.77 98.20 98.63 98.75
表 2  不同模型在UCSD数据集上的性能对比实验结果
模型 图片类别 P/% R/% Spe/% Acc/% NP/106 OAcc/%
Multiscale CNN[9] AMD 93.75 100 — 100 1.35 96.66
DME — — — —
正常眼底 100 93.33 — 93.33
Transfer Learning+SMA[10] AMD 100 100 100 100 — 99.89
DME 99.69 100 99.84 99.89
正常眼底 100 99.76 100 99.89
Resnet18[12] AMD 99.72 99.58 99.92 99.84 11.18 99.75
DME 99.80 99.72 99.92 99.84
正常眼底 99.70 99.84 99.78 99.80
MB-CNN AMD 100 99.7 100 99.2 1.35 99.92
DME 99.8 100 99.2 99.2
正常眼底 100 100 100 100
表 3  不同模型在Duke数据集上的对比实验结果
模型 图片类别 P/% R/% Spe/% Acc/% NP/106 OAcc/%
FPN+VGG16[11] — — — 96.5±0.8 — — 93.4±1.4
Resnet18[12] CNV 97.13 95.40 99.30 98.50 11.18 94.01±0.4
玻璃膜疣 91.47 88.77 96.73 94.37
正常眼底 94.30 96.53 93.70 95.16
MB-CNN CNV 96.8 96.1 99.2 98.6 1.35 94.37±0.4
玻璃膜疣 91.7 89.6 96.6 94.6
正常眼底 94.9 96.4 94.6 95.5
表 4  不同模型在NEH数据集上的实验对比结果
图 5  OCT热力图
1 FERRIS III F L, WILKINSON C P, BIRD A, et al Clinical classification of age-related macular degeneration[J]. Ophthalmology, 2013, 120 (4): 844- 851
doi: 10.1016/j.ophtha.2012.10.036
2 张昊瑞, 桂潇, 赵娜, 等 糖尿病性黄斑水肿的药物治疗研究进展[J]. 中国眼耳鼻喉科杂志, 2021, 21 (3): 226- 229
ZHANG Hao-rui, GUI Xiao, ZHAO Na, et al Research progress on pharmacotherapy for diabetic macular edema[J]. Chinese Journal of Ophthalmology and Otorhinolaryngology, 2021, 21 (3): 226- 229
3 HUANG D, SWANSON E A, LIN C P, et al Optical coherence tomography[J]. Science, 1991, 254 (5035): 1178- 1181
doi: 10.1126/science.1957169
4 PODOLEANU A G Optical coherence tomography[J]. Journal of Microscopy, 2012, 247 (3): 209- 219
doi: 10.1111/j.1365-2818.2012.03619.x
5 FANG L, WANG C, LI S, et al Attention to lesion: lesion-aware convolutional neural network for retinal optical coherence tomography image classification[J]. IEEE Transactions on Medical Imaging, 2019, 38 (8): 1959- 1970
doi: 10.1109/TMI.2019.2898414
6 LIU X, BAI Y, CAO J, et al Joint disease classification and lesion segmentation via one-stage attention-based convolutional neural network in OCT images[J]. Biomedical Signal Processing and Control, 2022, 71: 103087
doi: 10.1016/j.bspc.2021.103087
7 MISHRA S S, MANDAL B, PUHAN N B Perturbed composite attention model for macular optical coherence tomography image classification[J]. IEEE Transactions on Artificial Intelligence, 2022, 3 (4): 625- 635
doi: 10.1109/TAI.2021.3135797
8 DAS V, DANDAPAT S, BORA P K Automated classification of retinal OCT images using a deep multi-scale fusion CNN[J]. IEEE Sensors Journal, 2021, 21 (20): 23256- 23265
doi: 10.1109/JSEN.2021.3108642
9 THOMAS A, HARIKRISHNAN P M, KRISHAN A K, et al A novel multiscale convolutional neural network based age-related macular degeneration detection using OCT images[J]. Biomedical Signal Processing and Control, 2021, 67: 102538
doi: 10.1016/j.bspc.2021.102538
10 TOĞAÇAR M, ERGEN B, TÜMEN V Use of dominant activations obtained by processing OCT images with the CNNs and slime mold method in retinal disease detection[J]. Biocybernetics and Biomedical Engineering, 2022, 42 (2): 646- 666
doi: 10.1016/j.bbe.2022.05.005
11 SOTOUDEH-PAIMA S, JODEIRI A, HAJIZADEH F, et al Multi-scale convolutional neural network for automated AMD classification using retinal OCT images[J]. Computers in Biology and Medicine, 2022, 144: 105368
doi: 10.1016/j.compbiomed.2022.105368
12 HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition [C]// Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016: 770-778.
13 HE K, ZHANG X, REN S, et al. Identity mappings in deep residual networks [C]// Computer Vision–ECCV 2016. [S.l.]: Springer, 2016: 630-645.
14 LIU Z, MAO H, WU C, et al. A convnet for the 2020s [C]// Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New Orleans: IEEE, 2022: 11976-11986.
15 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. Boston: IEEE, 2015: 1-9.
16 CHOLLET F. Xception: deep learning with depthwise separable convolutions [C]// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu: IEEE, 2017: 1251-1258.
17 HAN K, WANG Y, TIAN Q, et al. GhostNet: more features from cheap operations [C]// Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE, 2020: 1580-1589.
18 IOFFE S, SZEGEDY C. Batch normalization: accelerating deep network training by reducing internal covariate shift[C]// Proceeding of the 32nd International Conference on Machine Learning.[S.l.]: PMLR, 2015: 448-456.
19 KERMANY D S, GOLDBAUM M, CAI W, et al Identifying medical diagnoses and treatable diseases by image-based deep learning[J]. Cell, 2018, 172 (5): 1122- 1131
doi: 10.1016/j.cell.2018.02.010
[1] 韩俊,袁小平,王准,陈烨. 基于YOLOv5s的无人机密集小目标检测算法[J]. 浙江大学学报(工学版), 2023, 57(6): 1224-1233.
[2] 艾青林,崔景瑞,吕冰海,童桐. 基于损伤区域融合变换的轴承鼓形滚子表面损伤检测方法[J]. 浙江大学学报(工学版), 2023, 57(5): 1009-1020.
[3] 张剑钊,郭继昌,汪昱东. 基于融合逆透射率图的水下图像增强算法[J]. 浙江大学学报(工学版), 2023, 57(5): 921-929.
[4] 杨帆,宁博,李怀清,周新,李冠宇. 基于语义增强特征融合的多模态图像检索模型[J]. 浙江大学学报(工学版), 2023, 57(2): 252-258.
[5] 陈巧红,孙佳锦,漏杨波,方志坚. 基于多任务学习与层叠 Transformer 的多模态情感分析模型[J]. 浙江大学学报(工学版), 2023, 57(12): 2421-2429.
[6] 艾青林,杨佳豪,崔景瑞. 基于自适应增殖数据增强与全局特征融合的小目标行人检测[J]. 浙江大学学报(工学版), 2023, 57(10): 1933-1944.
[7] 莫建文,李晋,蔡晓东,陈锦威. 基于门控特征融合与中心损失的目标识别[J]. 浙江大学学报(工学版), 2023, 57(10): 2011-2017.
[8] 吴兰,王涵,李斌全,李崇阳,孔凡士. 基于自监督任务的多源无监督域适应法[J]. 浙江大学学报(工学版), 2022, 56(4): 754-763.
[9] 张娜,戚旭磊,包晓安,吴彪,涂小妹,金瑜婷. 基于优化预测定位的单阶段目标检测算法[J]. 浙江大学学报(工学版), 2022, 56(4): 783-794.
[10] 黄新宇,游帆,张沛,张昭,张柏礼,吕建华,徐立臻. 基于多分类及特征融合的静默活体检测算法[J]. 浙江大学学报(工学版), 2022, 56(2): 263-270.
[11] 于楠晶,范晓飚,邓天民,冒国韬. 基于多头自注意力的复杂背景船舶检测算法[J]. 浙江大学学报(工学版), 2022, 56(12): 2392-2402.
[12] 李明,段立娟,王文健,恩擎. 基于显著稀疏强关联的脑功能连接分类方法[J]. 浙江大学学报(工学版), 2022, 56(11): 2232-2240.
[13] 徐泽鑫,段立娟,王文健,恩擎. 基于上下文特征融合的代码漏洞检测方法[J]. 浙江大学学报(工学版), 2022, 56(11): 2260-2270.
[14] 张云佐,郭威,蔡昭权,李文博. 联合多尺度与注意力机制的遥感图像目标检测[J]. 浙江大学学报(工学版), 2022, 56(11): 2215-2223.
[15] 杨栋杰,高贤君,冉树浩,张广斌,王萍,杨元维. 基于多重多尺度融合注意力网络的建筑物提取[J]. 浙江大学学报(工学版), 2022, 56(10): 1924-1934.