Turn off MathJax
Article Contents
Jing Zhang, Yi-Bo Huo, Jia-Liang Yang, Xiang-Zhou Wang, Bo-Yun Yan, Xiao-Hui Du, Ru-Qian Hao, Fang Yang, Juan-Xiu Liu, Lin Liu, Yong Liu, Hou-Bin Zhang. Automatic counting of retinal ganglion cells in the entire mouse retina based on improved YOLOv5. Zoological Research, 2022, 43(5): 738-749. doi: 10.24272/j.issn.2095-8137.2022.025
Citation: Jing Zhang, Yi-Bo Huo, Jia-Liang Yang, Xiang-Zhou Wang, Bo-Yun Yan, Xiao-Hui Du, Ru-Qian Hao, Fang Yang, Juan-Xiu Liu, Lin Liu, Yong Liu, Hou-Bin Zhang. Automatic counting of retinal ganglion cells in the entire mouse retina based on improved YOLOv5. Zoological Research, 2022, 43(5): 738-749. doi: 10.24272/j.issn.2095-8137.2022.025

Automatic counting of retinal ganglion cells in the entire mouse retina based on improved YOLOv5

doi: 10.24272/j.issn.2095-8137.2022.025
Funds:  This work was supported by the National Natural Science Foundation of China (61405028 to J.Z., 81770935 to H.B.Z.), Fundamental Research Funds for the Central Universities (University of Electronic Science and Technology of China) (ZYGX2019J053 to J.Z.), and Department of Science and Technology of Sichuan Province, China (2020YJ0445 to H.B.Z.)
More Information
  • Glaucoma is characterized by the progressive loss of retinal ganglion cells (RGCs), although the pathogenic mechanism remains largely unknown. To study the mechanism and assess RGC degradation, mouse models are often used to simulate human glaucoma and specific markers are used to label and quantify RGCs. However, manually counting RGCs is time-consuming and prone to distortion due to subjective bias. Furthermore, semi-automated counting methods can produce significant differences due to different parameters, thereby failing objective evaluation. Here, to improve counting accuracy and efficiency, we developed an automated algorithm based on the improved YOLOv5 model, which uses five channels instead of one, with a squeeze-and-excitation block added. The complete number of RGCs in an intact mouse retina was obtained by dividing the retina into small overlapping areas and counting, and then merging the divided areas using a non-maximum suppression algorithm. The automated quantification results showed very strong correlation (mean Pearson correlation coefficient of 0.993) with manual counting. Importantly, the model achieved an average precision of 0.981. Furthermore, the graphics processing unit (GPU) calculation time for each retina was less than 1 min. The developed software has been uploaded online as a free and convenient tool for studies using mouse models of glaucoma, which should help elucidate disease pathogenesis and potential therapeutics.
  • loading
  • [1]
    Abràmoff MD, Magalhães PJ, Ram SJ. 2004. Image processing with ImageJ. Biophotonics International, 11(7): 36−42.
    [2]
    Balendra SI, Normando EM, Bloom PA, Cordeiro MF. 2015. Advances in retinal ganglion cell imaging. Eye, 29(10): 1260−1269. doi: 10.1038/eye.2015.154
    [3]
    Berkelaar M, Clarke DB, Wang YC, Bray GM, Aguayo AJ. 1994. Axotomy results in delayed death and apoptosis of retinal ganglion cells in adult rats. The Journal of Neuroscience, 14(7): 4368−4374. doi: 10.1523/JNEUROSCI.14-07-04368.1994
    [4]
    Bochkovskiy A, Wang CY, Liao HYM. 2020. Yolov4: Optimal speed and accuracy of object detection. arXiv preprint arXiv: 2004.10934.
    [5]
    Casson RJ, Chidlow G, Wood JP, Crowston JG, Goldberg I. 2012. Definition of glaucoma: clinical and experimental concepts. Clinical & Experimental Ophthalmology, 40(4): 341−349.
    [6]
    Chen K, Wang JQ, Pang JM, Cao YH, Xiong Y, Li XX, et al. 2019. MMDetection: open mmlab detection toolbox and benchmark. arXiv preprint arXiv: 1906.07155.
    [7]
    Collins TJ. 2007. ImageJ for microscopy. Biotechniques, 43(S1): 25−30.
    [8]
    Cross T, Navarange R, Son JH, Burr W, Singh A, Zhang K, et al. 2020. Simple RGC: imageJ plugins for counting retinal ganglion cells and determining the transduction efficiency of viral vectors in retinal wholemounts. arXiv preprint arXiv: 2008.06276.
    [9]
    Dordea AC, Bray MA, Allen K, Logan DJ, Fei F, Malhotra R, et al. 2016. An open-source computational tool to automatically quantify immunolabeled retinal ganglion cells. Experimental Eye Research, 147: 50−56. doi: 10.1016/j.exer.2016.04.012
    [10]
    Evangelho K, Mogilevskaya M, Losada-Barragan M, Vargas-Sanchez JK. 2019. Pathophysiology of primary open-angle glaucoma from a neuroinflammatory and neurotoxicity perspective: a review of the literature. International Ophthalmology, 39(1): 259−271. doi: 10.1007/s10792-017-0795-9
    [11]
    Geeraerts E, Dekeyster E, Gaublomme D, Salinas-Navarro M, De Groef L, Moons L. 2016. A freely available semi-automated method for quantifying retinal ganglion cells in entire retinal flatmounts. Experimental Eye Research, 147: 105−113. doi: 10.1016/j.exer.2016.04.010
    [12]
    Gulcehre C, Moczulski M, Denil M, Bengio Y. 2016. Noisy Activation Functions. In: Proceedings of the 33rd International Conference on Machine Learning. New York: ACM, 3059–3068.
    [13]
    Guymer C, Damp L, Chidlow G, Wood J, Tang YF, Casson R. 2020. Software for quantifying and batch processing images of Brn3a and RBPMS immunolabelled retinal ganglion cells in retinal wholemounts. Translational Vision Science and Technology, 9(6): 28. doi: 10.1167/tvst.9.6.28
    [14]
    Howard AG, Zhu ML, Chen B, Kalenichenko D, Wang WJ, Weyand T, et al. 2017. Mobilenets: Efficient convolutional neural networks for mobile vision applications. arXiv preprint arXiv: 1704.04861.
    [15]
    Hu J, Shen L, Albanie S, Sun G, Wu EH. 2020. Squeeze-and-Excitation Networks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 42(8): 2011−2023. doi: 10.1109/TPAMI.2019.2913372
    [16]
    Ioffe S, Szegedy C. 2015. Batch normalization: Accelerating deep network training by reducing internal covariate shift. In: Proceedings of the 32nd International Conference on International Conference on Machine Learning. Lille: ACM, 448–456.
    [17]
    Jiang SM, Zeng LP, Zeng JH, Tang L, Chen XM, Wei X. 2015. β-III-Tubulin: a reliable marker for retinal ganglion cell labeling in experimental models of glaucoma. International Journal of Ophthalmology, 8(4): 643−652.
    [18]
    Jocher G, Nishimura K, Mineeva T, Vilariño R. 2020. YOLOv5.
    [19]
    Kuehn MH, Fingert JH, Kwon YH. 2005. Retinal ganglion cell death in glaucoma: mechanisms and neuroprotective strategies. Ophthalmology Clinics of North America, 18(3): 383−395. doi: 10.1016/j.ohc.2005.04.002
    [20]
    Kwon YH, Fingert JH, Kuehn MH, Alward WLM. 2009. Primary open-angle glaucoma. New England Journal of Medicine, 360(11): 1113−1124. doi: 10.1056/NEJMra0804630
    [21]
    Lam TT, Abler AS, Kwong JMK, Tso MOM. 1999. N-methyl-D-aspartate (NMDA)-induced apoptosis in rat retina. Investigative Ophthalmology and Visual Science, 40(10): 2391−2397.
    [22]
    Li Y, Schlamp CL, Nickells RW. 1999. Experimental induction of retinal ganglion cell death in adult mice. Investigative Ophthalmology & Visual Science, 40(5): 1004−1008.
    [23]
    Lin TY, Goyal P, Girshick R, He KM, Dollár P. 2017. Focal loss for dense object detection. In: Proceedings of the IEEE International Conference on Computer Vision. Venice: IEEE, 2999–3007.
    [24]
    Masin L, Claes M, Bergmans S, Cools L, Andries L, Davis BM, et al. 2021. A novel retinal ganglion cell quantification tool based on deep learning. Scientific Reports, 11(1): 702. doi: 10.1038/s41598-020-80308-y
    [25]
    Mead B, Tomarev S. 2016. Evaluating retinal ganglion cell loss and dysfunction. Experimental Eye Research, 151: 96−106. doi: 10.1016/j.exer.2016.08.006
    [26]
    Myles PS, Cui J. 2007. I. Using the Bland-Altman method to measure agreement with repeated measures. British Journal of Anaesthesia, 99(3): 309−311. doi: 10.1093/bja/aem214
    [27]
    Nadal-Nicolás FM, Jiménez-López M, Sobrado-Calvo P, Nieto-López L, Cánovas-Martínez I, Salinas-Navarro M, et al. 2009. Brn3a as a marker of retinal ganglion cells: qualitative and quantitative time course studies in naive and optic nerve-injured retinas. Investigative Ophthalmology & Visual Science, 50(8): 3860−3868.
    [28]
    Neubeck A, Van Gool L. 2006. Efficient non-maximum suppression. In: 18th International Conference on Pattern Recognition (ICPR 2006). Hong Kong, China: IEEE, 850–855.
    [29]
    Puth MT, Neuhäuser M, Ruxton GD. 2014. Effective use of Pearson's product-moment correlation coefficient. Animal Behaviour, 93: 183−189. doi: 10.1016/j.anbehav.2014.05.003
    [30]
    Ramachandran P, Zoph B, Le QV. 2018. Searching for activation functions. In: 6th International Conference on Learning Representations. Vancouver: ICLR.
    [31]
    Redmon J, Farhadi A. 2017. YOLO9000: Better, faster, stronger. In: 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Honolulu: IEEE, 6517–6525.
    [32]
    Redmon J, Farhadi A. 2018. Yolov3: an incremental improvement. arXiv preprint arXiv: 1804.02767.
    [33]
    Ren SQ, He KM, Girshick R, Sun J. 2017. Faster R-CNN: towards real-time object detection with region proposal networks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 39(6): 1137−1149. doi: 10.1109/TPAMI.2016.2577031
    [34]
    Rodriguez AR, De Sevilla Müller LP, Brecha NC. 2014. The RNA binding protein RBPMS is a selective marker of ganglion cells in the mammalian retina. Journal of Comparative Neurology, 522(6): 1411−1443. doi: 10.1002/cne.23521
    [35]
    Schmidhuber J. 2015. Deep learning in neural networks: an overview. Neural Networks, 61: 85−117. doi: 10.1016/j.neunet.2014.09.003
    [36]
    Van Etten A. 2018. You only look twice: Rapid multi-scale object detection in satellite imagery. arXiv preprint arXiv: 1805.09512.
    [37]
    Wang CY, Liao HYM, Wu YH, Chen PY, Hsieh JW, Yeh IH. 2020. CSPNet: a new backbone that can enhance learning capability of CNN. In: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW). Seattle: IEEE, 1571–1580.
    [38]
    Wang KX, Liew JH, Zou YT, Zhou DQ, Feng JS. 2019. Panet: few-shot image semantic segmentation with prototype alignment. In: Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision (ICCV). Seoul: IEEE, 9196–9205
    [39]
    Wilkinson L, Friendly M. 2009. The history of the cluster heat map. The American Statistician, 63(2): 179−184. doi: 10.1198/tas.2009.0033
    [40]
    Yang JL, Zou TD, Yang F, Yang ZL, Zhang HB. 2021. Inhibition of mTOR signaling by rapamycin protects photoreceptors from degeneration in rd1 mice. Zoological Research, 42(4): 482−486. (in Chinese) doi: 10.24272/j.issn.2095-8137.2021.049
  • ZR-2022-025-Supplementary Materials.pdf
  • 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(11)  / Tables(9)

    Article Metrics

    Article views (182) PDF downloads(51) Cited by()
    Proportional views
    Related

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return