Mingchao Li, Kun Huang, Xiao Ma, Yuexuan Wang, Wen Fan, Qiang Chen. Mask Distillation Network for Conjunctival Hyperemia Severity Classification. Machine Intelligence Research, vol. 20, no. 6, pp.909-922, 2023. https://doi.org/10.1007/s11633-022-1385-5
Citation: Mingchao Li, Kun Huang, Xiao Ma, Yuexuan Wang, Wen Fan, Qiang Chen. Mask Distillation Network for Conjunctival Hyperemia Severity Classification. Machine Intelligence Research, vol. 20, no. 6, pp.909-922, 2023. https://doi.org/10.1007/s11633-022-1385-5

Mask Distillation Network for Conjunctival Hyperemia Severity Classification

doi: 10.1007/s11633-022-1385-5
More Information
  • Author Bio:

    Mingchao Li received the B. Sc. degree in engineering mechanics from School of Computer Science and Engineering, Nanjing University of Science and Technology, China in 2016. He is currently a Ph.D. degree candidate in control science and engineering at School of Computer Science and Engineering, Nanjing University of Science and Technology, China. His research interests include medical image processing and deep learning. E-mail: chaosli@njust.edu.cn ORCID iD: 0000-0001-7034-3357

    Kun Huang received B. Sc. degree in computer science and technology from School of Computer Science and Technology, Zhejiang University of Technology, China in 2019. He is a Ph. D. degree candidate in control science and engineering at School of Computer Science and Engineering, Nanjing University of Science and Technology, China. His research interests include image generation and medical image processing. E-mail: 573172302@qq.com

    Xiao Ma received M. Sc. degree in automation from School of Computer Science and Engineering, Nanjing University of Science and Technology, China in 2021. He is a Ph. D. degree candidate in computer science and technology at School of Computer Science and Engineering, Nanjing University of Science and Technology, China. His research interests include weakly supervised learning and medical image processing. E-mail: maxiao@njust.edu.cn

    Yuexuan Wang received B. Sc. degree in electronic information science and technology from School of Physics and Information Technology, Shaanxi Normal University, China in 2017. She is a Ph. D. degree candidate in pattern recognition and intelligent system at School of Computer Science and Engineering from Nanjing University of Science and Technology, China. Her research interests include medical image processing and deep learning. E-mail: yxwang@njust.edu.cn

    Wen Fan received M. D. degree in clinical medicine from Wuhan University, China in 2012. She is currently an associate chief physician and associate professor with First Affiliated Hospital of Nanjing Medical University, China. Her research interests include retinal imaging and vitreoretinal diseases. E-mail: fanwen1029@163.com

    Qiang Chen received the B. Sc. degree in communication engineering and the Ph. D. degree in pattern recognition and intelligent system from Nanjing University of Science and Technology, China in 2002 and 2007, respectively. He held a post-doctoral position with Stanford University, USA from 2010 to 2011. He is currently a professor with Nanjing University of Science and Technology, China. His research interests include image processing and analysis, machine learning. E-mail: chen2qiang@njust.edu.cn (Corresponding author) ORCID iD: 0000-0002-6685-2447

  • Received Date: 2022-08-22
  • Accepted Date: 2022-10-20
  • Publish Date: 2023-12-01
  • To achieve automatic, fast and accurate severity classification of bulbar conjunctival hyperemia severity, we proposed a novel prior knowledge-based framework called mask distillation network (MDN). The proposed MDN consists of a segmentation network and a classification network with teacher-student branches. The segmentation network is used to generate a bulbar conjunctival mask and the classification network divides the severity of bulbar conjunctival hyperemia into four grades. In the classification network, we feed the original image and the image with the bulbar conjunctival mask into the student and teacher branches respectively, and an attention consistency loss and a classification consistency loss are used to keep a similar learning mode for these two branches. This design of “different input but same output”, named mask distillation (MD), aims to introduce the regional prior knowledge that “bulbar conjunctival hyperemia severity classification is only related to the bulbar conjunctiva region”. Extensive experiments on 5117 anterior segment images have proven the effectiveness of mask distillation technology: 1) The accuracy of the MDN student branch is 3.5% higher than that of a single optimal baseline network and 2% higher than that of the baseline network combination. 2) In the test phase, only the student branch is needed, and no additional segmentation network is required. The framework only takes 0.003 s to classify a single image, achieving the fastest speed in all the methods we compared. 3) Compared with a single baseline network, the attention of both teacher and student branches in the MDN has been intuitively improved.

     

  • loading
  • [1]
    P. J. Murphy, J. S. C. Lau, M. M. L. Sim, R. L. Woods. How red is a white eye? Clinical grading of normal conjunctival hyperaemia. Eye, vol. 21, no. 5, pp. 633–638, 2007. DOI: 10.1038/sj.eye.6702295.
    [2]
    A. van der Woerdt. Management of intraocular inflammatory disease. Clinical Techniques in Small Animal Practice, vol. 16, no. 1, pp. 58–61, 2001. DOI: 10.1053/svms.2001.22807.
    [3]
    M. Janssens. Efficacy of levocabastine in conjunctival provocation studies. Documenta Ophthalmologica, vol. 82, no. 4, pp. 341–351, 1992. DOI: 10.1007/BF00161022.
    [4]
    F. Honrubia, J. García-Sánchez, V. Polo, J. M. M. de la Casa, J. Soto. Conjunctival hyperaemia with the use of latanoprost versus other prostaglandin analogues in patients with ocular hypertension or glaucoma: A meta-analysis of randomised clinical trials. British Journal of Ophthalmology, vol. 93, no. 3, pp. 316–321, 2009. DOI: 10.1136/bjo.2007.135111.
    [5]
    D. S. Friedman, S. R. Hahn, L. Gelb, J. Tan, S. N. Shah, E. E. Kim, T. J. Zimmerman, H. A. Quigley. Doctor–patient communication, health-related beliefs, and adherence in glaucoma results from the glaucoma adherence and persistency study. Ophthalmology, vol. 15, no. 8, pp. 120–1327.e3, 2008. DOI: 10.1016/j.ophtha.2007.11.023.
    [6]
    T. Akagi, Y. Okamoto, T. Kameda, K. Suda, H. Nakanishi, M. Miyake, H. O. Ikeda, T. Yamada, S. Kadomoto, A. Uji, A. Tsujikawa. Short-term effects of different types of anti-glaucoma eyedrop on the sclero-conjunctival vasculature assessed using anterior segment OCTA in normal human eyes: A pilot study. Journal of Clinical Medicine, vol. 9, no. 2, Article number 4016, 2020. DOI: 10.3390/jcm9124016.
    [7]
    E. Terao, S. Nakakura, Y. Fujisawa, Y. Nagata, K. Ueda, Y. Kobayashi, S. Oogi, S. Dote, M. Shiraishi, H. Tabuchi, T. Yoneda, A. Fukushima, R. Asaoka, Y. Kiuchi. Time course of conjunctival hyperemia induced by omidenepag isopropyl ophthalmic solution 0.002%: A pilot, comparative study versus ripasudil 0.4%. BMJ Open Ophthalmology, vol. 5, no. 1, Article number e000538, 2020. DOI: 10.1136/bmjophth-2020-000538.
    [8]
    C. W. McMonnies, A. Chapman-Davies. Assessment of conjunctival hyperemia in contact lens wearers. Part I. American Academy of Optometry, vol. 64, no. 4, pp. 246–250, 1987. DOI: 10.1097/00006324-198704000-00003.
    [9]
    R. L. Terry, C. M. Schnider, B. A. Holden, R. Cornish, T. Grant, D. Sweeney, D. L. Hood, A. Back. CCLRU standards for success of daily and extended wear contact lenses. Optometry and Vision Science, vol. 70, no. 3, pp. 234–243, 1993. DOI: 10.1097/00006324-199303000-00011.
    [10]
    N. Efron. Grading scales for contact lens complications. Ophthalmic and Physiological Optics, vol. 18, no. 2, pp. 182–186, 1998. DOI: 10.1016/S0275-5408(97)00066-5.
    [11]
    M. M. Schulze, D. A. Jones, T. L. Simpson. The development of validated bulbar redness grading scales. Optometry and Vision Science, vol. 84, no. 10, pp. 976–983, 2007. DOI: 10.1097/OPX.0b013e318157ac9e.
    [12]
    W. C. Stewart, A. E. Kolker, J. A. Stewart, J. Leech, A. L. Jackson. Conjunctival hyperemia in healthy subjects after short-term dosing with latanoprost, bimatoprost, and travoprost. Americian Journal of Ophthalmology, vol. 135, no. 3, pp. 314–320, 2003. DOI: 10.1016/S0002-9394(02)01980-3.
    [13]
    M. R. H. M. Adnan, A. M. Zain, H. Haron, R. Alwee, M. Z. C. Azemin, A. O. Ibrahim. Eye redness image processing techniques. Journal of Physics:Conference Series, vol. 892, Article number 012019, 2017. DOI: 10.1088/1742-6596/892/1/012019.
    [14]
    M. M. Schulze, N. Hutchings, T. L. Simpson. Grading bulbar redness using cross-calibrated clinical grading scales. Investigative Ophthalmology &Visual Science, vol. 52, no. 8, pp. 5812–5817, 2011. DOI: 10.1167/iovs.10-7006.
    [15]
    T. Yoneda, T. Sumi, A. Takahashi, Y. Hoshikawa, M. Kobayashi, A. Fukushima. Automated hyperemia analysis software: Reliability and reproducibility in healthy subjects. Japanese Journal of Ophthalmology, vol. 56, no. 1, pp. 1–7, 2012. DOI: 10.1007/s10384-011-0107-2.
    [16]
    T. Sumi, T. Yoneda, K. Fukuda, Y. Hoshikawa, M. Kobayashi, M. Yanagi, Y. Kiuchi, K. Yasumitsu-Lovell, A. Fukushima. Development of automated conjunctival hyperemia analysis software. Cornea, vol. 32, no. 1, pp. S52–S59, 2013. DOI: 10.1097/ICO.0b013e3182a18e44.
    [17]
    B. Huntjens, M. Basi, M. Nagra. Evaluating a new objective grading software for conjunctival hyperaemia. Contact Lens and Anterior Eye, vol. 43, no. 2, pp. 137–143, 2020. DOI: 10.1016/j.clae.2019.07.003.
    [18]
    H. Masumoto, H. Tabuchi, T. Yoneda, S. Nakakura, H. Ohsugi, T. Sumi, A. Fukushima. Severity classification of conjunctival hyperaemia by deep neural network ensembles. Journal of Ophthalmology, vol. 2019, Article number 7820971, 2019. DOI: 10.1155/2019/7820971.
    [19]
    X. Q. Zhang, Y. Hu, Z. J. Xiao, J. S. Fang, R. Higashita, J. Liu. Machine learning for cataract classification/grading on ophthalmic imaging modalities: A survey. Machine Intelligence Research, vol. 19, no. 3, pp. 184–208, 2022. DOI: 10.1007/s11633-022-1329-0.
    [20]
    G. Hinton, O. Binyals, J. Dean. Distilling the knowledge in a neural network. Computer Science, vol. 14, no. 7, pp. 38–39, 2015.
    [21]
    S. Zagoruyko, N. Komodakis. Paying more attention to attention: Improving the performance of convolutional neural networks via attention transfer. In Proceedings of the 5th International Conference on Learning Representations, Toulon, France, 2017.
    [22]
    J. Yim, D. Joo, J. Bae, J. Kim. A gift from knowledge distillation: Fast optimization, network minimization and transfer learning. In Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, USA, pp. 7130–7138, 2017. DOI: 10.1109/CVPR.2017.754.
    [23]
    Z. H. Huang, N. Y. Wang. Like what you like: Knowledge distill via neuron selectivity transfer, [Online], Available: https://arxiv.org/abs/1707.01219, 2017.
    [24]
    S. H. Lin, H. W. Xie, B. Wang, K. C. Yu, X. J. Chang, X. D. Liang, G. Wang. Knowledge distillation via the target-aware transformer. In Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition, IEEE, New Orleans, USA, pp. 10905–10914, 2022. DOI: 10.1109/CVPR52688.2022.01064.
    [25]
    J. P. Gou, B. S. Yu, S. J. Maybank, D. C. Tao. Knowledge distillation: A survey. International Journal of Computer Vision, vol. 129, no. 6, pp. 1789–1819, 2021. DOI: 10.1007/s11263-021-01453-z.
    [26]
    H. Guo, K. Zheng, X. C. Fan, H. K. Yu, S. Wang. Visual attention consistency under image transforms for multi-label image classification. In Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition, IEEE, Long Beach, USA, pp. 729–739, 2019. DOI: 10.1109/CVPR.2019.00082.
    [27]
    B. L. Zhou, A. Khosla, A. Lapedriza, A. Oliva, A. Torralba. Learning deep features for discriminative localization. In Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, USA, pp. 2921–2929, 2016. DOI: 10.1109/CVPR.2016.319.
    [28]
    R. Ghosh, K. Ghosh, S. Maitra. Automatic detection and classification of diabetic retinopathy stages using CNN. In Proceedings of the 4th International Conference on Signal Processing and Integrated Networks, IEEE, Noida, India, pp. 550–554, 2017. DOI: 10.1109/SPIN.2017.8050011.
    [29]
    A. S. Krishnan, D. C. R, V. Bhat, P. B. Ramteke, S. G. Koolagudi. A transfer learning approach for diabetic retinopathy classification using deep convolutional neural networks. In Proceedings of the 15th IEEE India Council International Conference, Coimbatore, India, pp. 1–6, 2018. DOI: 10.1109/INDICON45594.2018.8987131.
    [30]
    R. R. Selvaraju, M. Cogswell, A. Das, R. Vedantam, D. Parikh, D. Batra. Grad-CAM: Visual explanations from deep networks via gradient-based localization. International Journal of Computer Vision, vol. 128, no. 2, pp. 336–359, 2020. DOI: 10.1007/s11263-019-01228-7.
    [31]
    K. Simonyan, A. Zisserman. Very deep convolutional networks for large-scale image recognition. In Proceedings of the 3rd International Conference on Learning Representations, San Diego, USA, 2015.
    [32]
    K. M. He, X. Y. Zhang, S. Q. Ren, J. Sun. Deep residual learning for image recognition. In Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, USA, pp. 770–778, 2015. DOI: 10.1109/CVPR.2016.90.
    [33]
    G. Huang, Z. Liu, L. van der Maaten, K. Q. Weinberger. Densely connected convolutional networks. In Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, USA, pp. 2261–2269, 2017. DOI: 10.1109/CVPR.2017.243.
    [34]
    C. Szegedy, V. Vanhoucke, S. Ioffe, J. Shlens, Z. Wojna. Rethinking the inception architecture for computer vision. In Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, USA, pp. 2818–2826, 2016. DOI: 10.1109/CVPR.2016.308.
    [35]
    F. Chollet. Xception: Deep learning with depthwise separable convolutions. In Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, USA, pp. 1800–1807, 2017. DOI: 10.1109/CVPR.2017.195.
    [36]
    M. Sandler, A. Howard, M. L. Zhu, A. Zhmoginov, L. C. Chen. MobileNetV2: Inverted residuals and linear bottlenecks. In Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition, IEEE, Salt Lake City, USA, pp. 4510–4520, 2018. DOI: 10.1109/CVPR.2018.00474.
    [37]
    J. Hu, L. Shen, G. Sun. “Squeeze-and-Excitation networks. In Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition, IEEE, Salt Lake City, USA, pp. 7132–7141, 2018. DOI: 10.1109/CVPR.2018.00745.
    [38]
    F. N. Iandola, S. Han, M. W. Moskewicz, K. Ashraf, W. J. Dally, K. Keutzer. SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and < 0.5 MB model size, [Online], Available: https://arxiv.org/abs/1602.07360, 2016.
    [39]
    M. X. Tan, B. Chen, R. M. Pang, V. Vasudevan, M. Sandler, A. Howard, Q. V. Le. MnasNet: Platform-aware neural architecture search for mobile. In Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition, IEEE, Long Beach, USA, pp. 2815–2823, 2019. DOI: 10.1109/CVPR.2019.00293.
  • 加载中

Catalog

    通讯作者: 陈斌, bchen63@163.com
    • 1. 

      沈阳化工大学材料科学与工程学院 沈阳 110142

    1. 本站搜索
    2. 百度学术搜索
    3. 万方数据库搜索
    4. CNKI搜索

    Figures(8)  / Tables(4)

    用微信扫码二维码

    分享至好友和朋友圈

    Article Metrics

    Article views (103) PDF downloads(2) Cited by()
    Proportional views
    Related

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return