kaz29Вестник Казахстанско-Британского технического университетаHerald of the Kazakh-British Technical University1998-66882959-8109Казахстанско-Британский Технический Университет10.55452/1998-6688-2021-18-3-75-82kaz29-105Research ArticleФИЗИКО-МАТЕМАТИЧЕСКИЕ И ТЕХНИЧЕСКИЕ НАУКИPHYSICAL, MATHEMATICAL AND TECHNICAL SCIENCESСРАВНЕНИЕ МОДЕЛЕЙ КОМПЬЮТЕРНОГО ЗРЕНИЯCOMPUTER VISION MODEL COMPARISONНамД.NamD.

050000, Алматы

Nam Diana

050000, Almaty

d_nam@kbtu.kzСавинаТ.SavinaT.

050000, Алматы

Savina Tamara

050000, Almaty

t_savina@kbtu.kzКазахстанско-Британский технический университетКазахстанKazakh-British Technical UniversityKazakhstan2021051120211837582Copyright © Нам Д., Савина Т., 20212021Нам Д., Савина Т.Nam D., Savina T.Данная работа распространяется под лицензией Creative Commons Attribution 4.0.This work is licensed under a Creative Commons Attribution 4.0 License.https://vestnik.kbtu.edu.kz/jour/article/view/105

Использование машинного обучения в области медицины является одной из самых сложных и досконально нерешенной задачей. В настоящее время существует множество различных алгоритмов для решения задач в области диагностики и сегментации биомедицинских изображений. Исследователи часто сталкиваются с проблемой выбора наилучшего метода, применимого к исследуемым данным. Мы провели эмпирическое исследование и сравнили 5 алгоритмов, которые способны решить задачу определения аномалии на медицинских снимках: R-CNN, Fast-RCNN, Faster-RCNN, Mask R CNN, U-Net, и Residual Neural Network. Преимущества автоматической обработки медицинских снимков очевидны: болезнь можно диагностировать быстрее, врачи получают удобный программный инструмент, а также снижается процент ошибок при обработке данных. Была поставлена задача – изучить, а в дальнейшем отобрать алгоритмы для дальнейшего тестирования на реальных данных. Отбор и изучение алгоритмов происходил на основе статей, описывающих архитектуру и применение алгоритмов компьютерного зрения.

The use of machine learning in the medical field is one of the most difficult and thoroughly unsolved problems. Currently, there are many different algorithms for solving problems in the field of diagnostics and segmentation of biomedical images. Researchers are often faced with the challenge of choosing the best method to apply towards their data. We conducted the empirical research and compared 5 algorithms that able to detect anomalies in the medical images: RCNN, Fast-RCNN, Faster-RCNN, Mask R CNN, U-Net, and Residual Neural Network. The advantages of automatic processing of the medical images are apparent: doctors get a convenient software tool that allows them to diagnose the disease faster and reduce possible errors. The task is to study and then select algorithms for further testing on the actual data. The selection and study of algorithms were based on articles describing the architecture and application of computer vision algorithms.

машинное обучениеглубокое обучениенейронные сетисверточные нейронные сетиmachine learningdeep learningneural networksconvolutional neural networksReferencesR. Girshick, J. Donahue, T. Darrell, and J. Malik, “Rich feature hierar-chies for accurate object detection and semantic segmentation,”Proceed-ings of the IEEE Computer Society Conference on Computer Vision andPattern Recognition, pp. 580–587, 2014.R. Girshick, J. Donahue, T. Darrell, and J. Malik, “Rich feature hierar-chies for accurate object detection and semantic segmentation,”Proceed-ings of the IEEE Computer Society Conference on Computer Vision andPattern Recognition, pp. 580–587, 2014.R. Girshick, “Fast R-CNN,”Proceedings of the IEEE International Conference on Computer Vision, vol. 2015 Inter, pp. 1440–1448, 2015R. Girshick, “Fast R-CNN,”Proceedings of the IEEE International Conference on Computer Vision, vol. 2015 Inter, pp. 1440–1448, 2015S. Ren, K. He, R. Girshick, and J. Sun, “Faster r-cnn: Towards real-timeobject detection with region proposal networks,” 2016.S. Ren, K. He, R. Girshick, and J. Sun, “Faster r-cnn: Towards real-timeobject detection with region proposal networks,” 2016.K. He, G. Gkioxari, P. Doll ár, and R. Girshick, “Mask R-CNN,”IEEETransactions on Pattern Analysis and Machine Intelligence, vol. 42, no. 2,pp. 386–397, 2020.K. He, G. Gkioxari, P. Doll ár, and R. Girshick, “Mask R-CNN,”IEEETransactions on Pattern Analysis and Machine Intelligence, vol. 42, no. 2,pp. 386–397, 2020.J. W. Johnson, “Adapting Mask-RCNN for automatic nucleus segmentation,” arXiv, pp. 1–7, 2018.J. W. Johnson, “Adapting Mask-RCNN for automatic nucleus segmentation,” arXiv, pp. 1–7, 2018.Image set BBBC038v1, available from the Broad Bioimage Benchmark Collection [Caicedo et al., Nature Methods, 2019]Image set BBBC038v1, available from the Broad Bioimage Benchmark Collection [Caicedo et al., Nature Methods, 2019]S. Liu, L. Qi, H. Qin, J. Shi, and J. Jia, “Path aggregation network forinstance segmentation,” 2018S. Liu, L. Qi, H. Qin, J. Shi, and J. Jia, “Path aggregation network forinstance segmentation,” 2018E. Kopelowitz and G. Engelhard, “Lung nodules detection and segmen-tation using 3d mask-rcnn,” 2019E. Kopelowitz and G. Engelhard, “Lung nodules detection and segmen-tation using 3d mask-rcnn,” 2019C. Zhao, J. Han, Y. Jia, and F. Gou, “Lung nodule detection via3D U-net and contextual convolutional neural network,”Proceedings -2018 International Conference on Networking and Network Applications,NaNA 2018, pp. 356–361, 2019. A. Pak, A. Ziyaden, K. Tukeshev, A. Jaxylykova, and D. Abdullina,“Comparative analysis of deep learning methods of detection of diabeticretinopathy,”Cogent Engineering, vol. 7, no. 1, p. 1805144, 2020.C. Zhao, J. Han, Y. Jia, and F. Gou, “Lung nodule detection via3D U-net and contextual convolutional neural network,”Proceedings -2018 International Conference on Networking and Network Applications,NaNA 2018, pp. 356–361, 2019. A. Pak, A. Ziyaden, K. Tukeshev, A. Jaxylykova, and D. Abdullina,“Comparative analysis of deep learning methods of detection of diabeticretinopathy,”Cogent Engineering, vol. 7, no. 1, p. 1805144, 2020.O. Ronneberger, P. Fischer, T. Brox, “U-net: Convolutional networks for biomedical image segmentation”, vol. 9351, pp.234-241, 2015.O. Ronneberger, P. Fischer, T. Brox, “U-net: Convolutional networks for biomedical image segmentation”, vol. 9351, pp.234-241, 2015.M. Zhou, M. Rahman, N. Siddiquee, Tajbakhsh, and J. Liang, Unet++, “A nested u-net architecture for medical image segmentation”, Springer International Publishing, 2018, vol. 11045 LNCS. [Online].M. Zhou, M. Rahman, N. Siddiquee, Tajbakhsh, and J. Liang, Unet++, “A nested u-net architecture for medical image segmentation”, Springer International Publishing, 2018, vol. 11045 LNCS. [Online].K.He, X.Zhang, S. Ren, “Deep residual learning for image recognition”, Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, vol. 2016-decem, p.770-778, 2016.K.He, X.Zhang, S. Ren, “Deep residual learning for image recognition”, Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, vol. 2016-decem, p.770-778, 2016.Romera-Paredes and P. H. S. Torr, “Recurrent instance segmentation,” Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 9910 LNCS, pp. 312–329, 2016.Romera-Paredes and P. H. S. Torr, “Recurrent instance segmentation,” Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 9910 LNCS, pp. 312–329, 2016.O. Vinyals, Q.Le, “A Neural Conversational Model”, vol. 37, 2016.O. Vinyals, Q.Le, “A Neural Conversational Model”, vol. 37, 2016.

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