sapiСистемный анализ и прикладная информатика«System analysis and applied information science»2309-49232414-0481Belarusian National Technical University10.21122/2309-4923-2023-1-25-34sapi-604Research ArticleОбработка информации и принятие решенийData processing and decision–makingНейросетевой классификатор для определения диабетической ретинопатии по изображениям сетчаткиA neural network classifier for detecting diabetic retinopathy from retinal imagesЛукашевичМ. М.LukashevichM. M.

Лукашевич Марина Михайловна - кандидат технических наук, доцент, докторант 

Минск, Республика Беларусь

PhD, Associate Professor, Postdoctoral Fellow

Minsk

lukashevich@bsuir.byБелорусский государственный университет информатики и радиоэлектроникиБеларусьBelarusian State University of Informatics and RadioelectronicsBelarus202301082023012534Copyright © Лукашевич М.М., 20232023Лукашевич М.М.Lukashevich M.M.Данная работа распространяется под лицензией Creative Commons Attribution 4.0.This work is licensed under a Creative Commons Attribution 4.0 License.https://sapi.bntu.by/jour/article/view/604

С распространением сахарного диабета диабетическая ретинопатия (ДР) становится основной проблемой здравоохранения (особенно в развивающихся странах). Долгосрочные осложнения, возникающие врезультате ДР, оказываютзначительноевлияниенапациентов. Ранняядиагностикаипоследующеелечение могут уменьшить наносимый здоровью вред. Предсказательная аналитика может быть основа на анализе изображений сетчатки глаза человека с помощью сверточных нейронных сетей. В работе исследования сосредоточены на разработке эффективного метода определения ДР на основе сверточной нейронной сети EfficientNet, технологии самообучения и операций аугментации данных. В результате экспериментов разработан нейросетевой классификатор на основе сверточных нейросетей, даны рекомендации для проведения аугментации данных. Эксперименты выполнены на публичном наборе данных и показали, что возможно достичь доли правильно классифицированных объектов равной 97,14 % на тестовом наборе из публичного набора данных.

With the spread of diabetes mellitus, diabetic retinopathy (DR) is becoming a major public health problem (especially in developing countries). The long-term complications resulting from DR have a significant impact on patients. Early diagnosis and subsequent treatment can reduce the damage to health. Predictive analytics can be based on the analysis of human retinal images using convolutional neural networks. In this paper, the research focuses on the development of an efficient method for DR detection based on the EfficientNet convolutional neural network, self-learning technology and data augmentation operations. As a result of the experiments, a neural network classifier based on convolutional neural networks is developed, recommendations for data augmentation operations are given. Experiments were performed on the public dataset and showed that it is possible to achieve the proportion of correctly classified objects equal to 97.14 % on the test set from the public dataset.

изображениедиабетическая ретинопатияобработка изображенийсверточные нейронные сетиimagediabetic retinopathyimage processingconvolutional neural networksРабота частично выполнена в рамках проекта БРФФИ Ф21ПАКГ 001.ReferencesBourne, R.R. Vision Loss Expert Group. Causes of vision loss worldwide, 1990-2010: a systematic analysis / R.R. Bourne, G.A. Stevens, R.A. White, J.L. Smith, S.R. Flaxman, H. Price, J.B. Jonas, J. Keeffe, J. Leasher, K. Naidoo, K. Pesudovs, S. Resnikoff, H.R. Taylor // Lancet Glob Health 2013. – T. 1, № 6. ‒ C. 339-349. doi: 10.1016/S2214-109X(13)70113-XBourne, R.R. Vision Loss Expert Group. Causes of vision loss worldwide, 1990-2010: a systematic analysis / R.R. Bourne, G.A. Stevens, R.A. White, J.L. Smith, S.R. Flaxman, H. Price, J.B. Jonas, J. Keeffe, J. Leasher, K. Naidoo, K. Pesudovs, S. Resnikoff, H.R. Taylor // Lancet Glob Health ‒ 2013. – T. 1, № 6. ‒ C. 339-349. doi: 10.1016/S2214-109X(13)70113-XTaylor, R., Batey, D. Handbook of retinal screening in diabetes. ‒ Wiley, 2006.Taylor, R., Batey, D. Handbook of retinal screening in diabetes. Wiley (2006).The Diabetic Retinopathy Clinical Research Network: Aflibercept, bevacizumab, or ranibizumab for diabetic macular edema. N Engl J Med372(13):1193-1203, 2015. doi: 10.1056/NEJMoa1414264The Diabetic Retinopathy Clinical Research Network: Aflibercept, bevacizumab, or ranibizumab for diabetic macular edema. N Engl J Med372(13):1193-1203, 2015. doi: 10.1056/NEJMoa1414264Beaulieu, W.T., Bressler, N.M., Melia, M., et al. Panretinal photocoagulation versus ranibizumab for proliferative diabetic retinopathy: Patient-centered outcomes from a randomized clinical trial. Am J Ophthalmol 170:206-213, 2016. doi: 10.1016/j.ajo.2016.08.008Beaulieu, W.T., Bressler, N.M., Melia, M., et al. Panretinal photocoagulation versus ranibizumab for proliferative diabetic retinopathy: Patient-centered outcomes from a randomized clinical trial. Am J Ophthalmol 170:206-213, 2016. doi: 10.1016/j.ajo.2016.08.008Mookiah, M.R.K., Acharya, U.R., Chua, C.K., Lim, C.M., Ng, E., Laude, A. Computer-aided diagnosis of diabetic retinopathy: a review. Comput Biol Med. 2013;43(12):2136–2155. doi: 10.1016/j.compbiomed.2013.10.007Mookiah, M.R.K., Acharya, U.R., Chua, C.K., Lim, C.M., Ng, E., Laude, A. Computer-aided diagnosis of diabetic retinopathy: a review. Comput Biol Med. 2013;43(12):2136–2155. doi: 10.1016/j.compbiomed.2013.10.007Lam, C., Yi, D., Guo, M., Lindsey, T. Automated detection of diabetic retinopathy using deep learning. AMIA Summit Translat Sci Proc. 2018;2018:147.Lam, C., Yi, D., Guo, M., Lindsey, T. Automated detection of diabetic retinopathy using deep learning. AMIA Summit Translat Sci Proc. 2018;2018:147.Bargshady, G., Zhou, X., Deo, R., Soar, J., Whittaker, F., Wang, H. Enhanced deep learning algorithm development to detect pain intensity from facial expression images. Expert Syst Appl. 2020;149:113305. doi: 10.1016/j.eswa.2020.113305Bargshady, G., Zhou, X., Deo, R., Soar, J., Whittaker, F., Wang, H. Enhanced deep learning algorithm development to detect pain intensity from facial expression images. Expert Syst Appl. 2020;149:113305. doi: 10.1016/j.eswa.2020.113305Sarki, R., Ahmed, K., Wang, H., Zhang, Y. Automatic detection of diabetic eye disease through deep learning using fundus images: a survey. IEEE Access. 2020;8:151133–151149. doi: 10.1109/ACCESS.2020.3015258Sarki, R., Ahmed, K., Wang, H., Zhang, Y. Automatic detection of diabetic eye disease through deep learning using fundus images: a survey. IEEE Access. 2020;8:151133–151149. doi: 10.1109/ACCESS.2020.3015258Gargeya, R., Leng, T. Automated identification of diabetic retinopathy using deep learning. Ophthalmology. 2017;124(7):962–969. doi: 10.1016/j.ophtha.2017.02.008Gargeya, R., Leng, T. Automated identification of diabetic retinopathy using deep learning. Ophthalmology. 2017;124(7):962–969. doi: 10.1016/j.ophtha.2017.02.008Litjens, G., Kooi ,T., Bejnordi, B.E., Setio, A.A.A., Ciompi, F., Ghafoorian, M., Van Der Laak, J.A., Van Ginneken, B., Sánchez, C.I. A survey on deep learning in medical image analysis. Med Image Anal. 2017;42:60–88. doi: 10.1016/j.media.2017.07.005Litjens, G., Kooi, T., Bejnordi, B.E., Setio, A.A.A., Ciompi, F., Ghafoorian, M., Van Der Laak, J.A., Van Ginneken, B., Sánchez, C.I. A survey on deep learning in medical image analysis. Med Image Anal. 2017;42:60–88. doi: 10.1016/j.media.2017.07.005EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks [Электронный ресурс]. – Режим доступа: https://arxiv.org/abs/1905.11946EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks [Electronic resource]. – Avialable: https://arxiv.org/abs/1905.11946Wang, Jing & Yang, Liu & Huo, Zhanqiang & He, Weifeng & Luo, Junwei. (2020). Multi-Label Classification of Fundus Images With EfficientNet. IEEE Access. 8. 212499-212508. doi: 10.1109/ACCESS.2020.3040275.Wang, Jing & Yang, Liu & Huo, Zhanqiang & He, Weifeng & Luo, Junwei (2020). Multi-Label Classification of Fundus Images With EfficientNet. IEEE Access. 8. 212499-212508. doi: 10.1109/ACCESS.2020.3040275Parthiban, K., Kamarasan, M. EfficientNet with Optimal Wavelet Neural Network for DR Detection and Grading, 2022 4th International Conference on Smart Systems and Inventive Technology (ICSSIT), Tirunelveli, India, 2022, pp. 10811086, doi: 10.1109/ICSSIT53264.2022.9716528Parthiban, K., Kamarasan, M. EfficientNet with Optimal Wavelet Neural Network for DR Detection and Grading, 2022 4th International Conference on Smart Systems and Inventive Technology (ICSSIT), Tirunelveli, India, 2022, pp. 10811086, doi: 10.1109/ICSSIT53264.2022.9716528Bulut, B., Kalın, V. , Bektaş Güneş, B. & Khazhin, R. (2022). Classification of Eye Disease from Fundus Images Using EfficientNet. Artificial Intelligence Theory and Applications , 2(1), 1-7 . Retrieved from https://dergipark.org.tr/en/pub/aita/issue/70443/1134144Bulut, B., Kalın, V., Bektaş Güneş, B. & Khazhin, R. (2022). Classification of Eye Disease from Fundus Images Using EfficientNet. Artificial Intelligence Theory and Applications , 2 (1), 1-7. Retrieved from https://dergipark.org.tr/en/pub/aita/ issue/70443/1134144Zhu, S., Lu B., Wang, C, Wu, M., Zheng, B., Jiang, Q., Wei, R., Cao, Q., Yang, W. Screening of Common Retinal Diseases Using Six-Category Models Based on EfficientNet. Front Med (Lausanne). 2022 Feb 23;9:808402. doi: 10.3389/ fmed.2022.808402. PMID: 35280876; PMCID: PMC8904395.Zhu, S., Lu B., Wang, C, Wu, M., Zheng, B., Jiang, Q., Wei, R., Cao, Q., Yang, W. Screening of Common Retinal Diseases Using Six-Category Models Based on EfficientNet. Front Med (Lausanne). 2022 Feb 23;9:808402. doi: 10.3389/fmed.2022.808402. PMID: 35280876; PMCID: PMC8904395.ImageNet [Электронный ресурс]. – Режим доступа: https://www.image-net.org/. – Дата доступа : 28.08.2023.ImageNet [Electronic resource]. – Avialable: https://www.image-net.org/. – Date of access : 28.08.2023.Deep transfer learning for image classification: a survey [Электронный ресурс]. – Режим доступа: https://arxiv. org/abs/2205.09904. – Дата доступа : 28.08.2023.Deep transfer learning for image classification: a survey [Electronic resource]. – Avialable: https://arxiv.org/abs/2205.09904. – Date of access : 28.08.2023.Kim, H.E., Cosa-Linan, A., Santhanam, N. et al. Transfer learning for medical image classification: a literature review. BMC Med Imaging 22, 69 (2022). doi: 10.1186/s12880-022-00793-7Kim, H.E., Cosa-Linan, A., Santhanam, N. et al. Transfer learning for medical image classification: a literature review. BMC Med Imaging 22, 69 (2022). doi: 10.1186/s12880-022-00793-7Zhou, Z., Shin, J., Zhang, L., Gurudu, S., Gotway, M., Liang, J. Fine-tuning Convolutional Neural Networks for Biomedical Image Analysis: Actively and Incrementally. Proc IEEE Comput Soc Conf Comput Vis Pattern Recognit. 2017 Jul;2017:4761-4772. doi: 10.1109/CVPR.2017.506. Epub 2017 Nov 9. PMID: 30337799; PMCID: PMC6191179.Zhou, Z., Shin, J., Zhang, L., Gurudu, S., Gotway, M., Liang, J. Fine-tuning Convolutional Neural Networks for Biomedical Image Analysis: Actively and Incrementally. Proc IEEE Comput Soc Conf Comput Vis Pattern Recognit. 2017 Jul;2017:4761-4772. doi: 10.1109/CVPR.2017.506. Epub 2017 Nov 9. PMID: 30337799; PMCID: PMC6191179.COCO – Common Objects in Context [Электронный ресурс]. – Режим доступа: https://cocodataset.org/. – Дата доступа : 28.08.2023.COCO – Common Objects in Context [Electronic resource]. – Avialable: https://cocodataset.org/. – Date of access : 28.08.2023.Rethinking ImageNet Pre-training [Электронный ресурс]. – Режим доступа: https://arxiv.org/pdf/1811.08883.pdf. – Дата доступа : 28.08.2023.Rethinking ImageNet Pre-training [Electronic resource]. – Avialable: https://arxiv.org/pdf/1811.08883.pdf. – Date of access : 28.08.2023.Rethinking Pre-training and Self-training [Электронный ресурс]. – Режим доступа: https://arxiv.org/pdf/2006.06882.pdf. – Дата доступа : 28.08.2023.Rethinking Pre-training and Self-training [Electronic resource]. – Avialable: https://arxiv.org/pdf/2006.06882. pdf. – Date of access : 28.08.2023.Self-training with Noisy Student improves ImageNet classification [Электронный ресурс]. – Режим доступа: https://arxiv.org/pdf/1911.04252.pdf. – Дата доступа : 28.08.2023.Self-training with Noisy Student improves ImageNet classification [Electronic resource]. – Avialable: https://arxiv. org/pdf/1911.04252.pdf. – Date of access : 28.08.2023.Models [Электронный ресурс]. – Режим доступа: https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/tf_efficientnet_b4_ns-d6313a46.pth. – Дата доступа : 28.08.2023.Models [Electronic resource]. – Avialable: https://github.com/rwightman/pytorch-image-models/releases/download/ v0.1-weights/tf_efficientnet_b4_ns-d6313a46.pth. – Date of access : 28.08.2023.APTOS 2019 Blindness Detection [Электронный ресурс]. – Режим доступа: https://www.kaggle.com/c/aptos2019-blindness-detection/. – Дата доступа : 28.08.2023.APTOS 2019 Blindness Detection [Electronic resource]. – Avialable: https://www.kaggle.com/c/aptos2019blindness-detection/. – Date of access : 28.08.2023.PyTorch [Электронный ресурс]. – Режим доступа: https://pytorch.org. – Дата доступа : 28.08.2023.PyTorch [Electronic resource]. – Avialable: https://pytorch.org. – Date of access : 28.08.2023.Image Data Augmentation for Deep Learning: A Survey [Электронный ресурс]. – Режим доступа: https://arxiv.org/abs/2204.08610. – Дата доступа : 28.08.2023.Image Data Augmentation for Deep Learning: A Survey [Electronic resource]. – Avialable: https://arxiv.org/abs/2204.08610. – Date of access : 28.08.2023.Adam: A Method for Stochastic Optimization [Электронный ресурс]. – Режим доступа: https://arxiv.org/abs/1412.6980. – Дата доступа : 28.08.2023.Adam: A Method for Stochastic Optimization [Electronic resource]. – Avialable: https://arxiv.org/abs/1412.6980. – Date of access : 28.08.2023.Uses and Abuses of the Cross-Entropy Loss: Case Studies in Modern Deep Learning [Электронный ресурс]. – Режим доступа: https://arxiv.org/abs/2011.05231. – Дата доступа : 28.08.2023.Uses and Abuses of the Cross-Entropy Loss: Case Studies in Modern Deep Learning [Electronic resource]. – Avialable: https://arxiv.org/abs/2011.05231. – Date of access : 28.08.2023.

The authors declare that there are no conflicts of interest present.