• Laser & Optoelectronics Progress
  • Vol. 58, Issue 1, 117002 (2021)
Lian Chaoming1, Zhong Shuncong1、*, Zhang Tianfu1, Zhou Ning1, and Xie Maosong2
Author Affiliations
  • 1School of Mechanical Engineering and Automation, Fuzhou University, Fuzhou, Fujian 350108, China
  • 2The First Affiliated Hospital of Fujian Medical University, Fuzhou, Fujian 350005, China
  • show less
    DOI: 10.3788/LOP202158.0117002 Cite this Article Set citation alerts
    Lian Chaoming, Zhong Shuncong, Zhang Tianfu, Zhou Ning, Xie Maosong. Transfer Learning-Based Classification of Optical Coherence Tomography Retinal Images[J]. Laser & Optoelectronics Progress, 2021, 58(1): 117002 Copy Citation Text show less
    Global average pooling function for image feature extraction
    Fig. 1. Global average pooling function for image feature extraction
    OCT images of diabetic retinopathy. (a) Choroidal neovascularization; (b) diabetic macular edema; (c) drusen; (d) normal retina
    Fig. 2. OCT images of diabetic retinopathy. (a) Choroidal neovascularization; (b) diabetic macular edema; (c) drusen; (d) normal retina
    Images before and after denoising
    Fig. 3. Images before and after denoising
    GAP transfer learning network model structure
    Fig. 4. GAP transfer learning network model structure
    Training set curves of control experimental networks. (a)(d) Training set curves of direct transfer learning group; (b)(e) training set curves of fine-tuning transfer learning group; (c)(f) training set curves of GAP transfer learning group
    Fig. 5. Training set curves of control experimental networks. (a)(d) Training set curves of direct transfer learning group; (b)(e) training set curves of fine-tuning transfer learning group; (c)(f) training set curves of GAP transfer learning group
    ROC curve of GAP transfer learning group network classification result. (a) InceptionV3 network; (b) VGG19 network; (c) ResNet50 network
    Fig. 6. ROC curve of GAP transfer learning group network classification result. (a) InceptionV3 network; (b) VGG19 network; (c) ResNet50 network
    CategoryNumber of images in data set
    TrainingValidationTestAll
    CNV324773820191238209
    DME9647113656711350
    Drusen73238624318616
    Normal223722632131626320
    All718198450422684495
    Table 1. Retina data set partition
    GroupTest accuracy/%
    InceptionV3VGG19ResNet50
    Fine-tuning transferlearning group94.994.892.9
    GAP transferlearning group97.396.994.4
    Table 2. Test accuracy of different pre-training models in control experiment
    Lian Chaoming, Zhong Shuncong, Zhang Tianfu, Zhou Ning, Xie Maosong. Transfer Learning-Based Classification of Optical Coherence Tomography Retinal Images[J]. Laser & Optoelectronics Progress, 2021, 58(1): 117002
    Download Citation