kaz29Вестник Казахстанско-Британского технического университетаHerald of the Kazakh-British Technical University1998-66882959-8109Казахстанско-Британский Технический Университет10.55452/1998-6688-2026-23-1-147-162kaz29-2509Research ArticleКОМПЬЮТЕРНЫЕ НАУКИCOMPUTER SCIENCEАНАЛИЗ РЕНТГЕНОВСКИХ СНИМКОВ ЛЕГОЧНОЙ СИСТЕМЫ ЧЕЛОВЕКА МЕТОДОМ МАШИННОГО ОБУЧЕНИЯ С ЦЕЛЬЮ ПОСТАНОВКИ ПРЕДВАРИТЕЛЬНОГО ДИАГНОЗАMACHINE LEARNING ANALYSIS OF HUMAN LUNG X-RAY IMAGES TO MAKE A PRELIMINARY DIAGNOSIShttps://orcid.org/0000-0002-1937-8615ИсаховA. А.IssakhovA. A.

профессор

г. Алматы

Professor

Almaty

alibek.issakhov@gmail.comhttps://orcid.org/0000-0002-5967-6959АбылкасымоваА. Б.AbylkassymovaA. B.

ассоциированный профессор

г. Алматы

Associate Professor

Almaty

abylkassymova.aizhan@gmail.comhttps://orcid.org/0009-0004-5395-4569БревновТ.BrevnovT.

бакалавр

г. Алматы

Bachelor’s student

Almaty

t_brevnov@kbtu.kzКазахстанско-Британский технический университетКазахстанKazakh-British technical universityKazakhstan202629032026231147162Copyright © Исахов A.А., Абылкасымова А.Б., Бревнов Т., 20262026Исахов A.А., Абылкасымова А.Б., Бревнов Т.Issakhov A.A., Abylkassymova A.B., Brevnov 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/2509

В статье представлено комплексное исследование применения методов машинного обучения для автоматизированного анализа рентгенографических изображений органов дыхательной системы с целью раннего выявления патологических изменений. Предложена и реализована методика классификации легочных заболеваний на основе ансамбля глубоких сверточных нейронных сетей, включающего архитектуры DenseNet121, MobileNetV2, EfficientNetB0, SENet и ShuffleNetV2. В рамках исследования проведен сравнительный анализ эффективности различных методов предобработки изображений, включая использование исходных черно-белых рентгеновских снимков без дополнительной обработки, применение метода CLAHE (Contrast Limited Adaptive Histogram Equalization) в сочетании с цветовой фильтрацией, а также использование нейросетевого денойзера DynamicCNN для подавления шумов. Экспериментальные результаты показали, что ансамблевый подход с применением стратегии мягкого голосования (soft voting) обеспечивает статистически значимое повышение точности классификации по сравнению с отдельными моделями. Полученные результаты подтверждают высокую эффективность предложенного подхода и демонстрируют перспективность использования ансамблевых моделей глубокого обучения в задачах медицинской диагностики и поддержки принятия клинических решений.

The article presents a comprehensive study of the application of machine learning methods for automated analysis of radiographic images of the respiratory system for the early detection of pathological changes. A method for classifying pulmonary diseases based on an ensemble of deep convolutional neural networks, including the DenseNet121, MobileNetV2, EfficientNetB0, SENet, and ShuffleNetV2 architectures, is proposed and implemented. The study included a comparative analysis of the effectiveness of various image preprocessing methods, including the use of raw black-and-white X-ray images without additional processing, the use of the CLAHE (Contrast Limited Adaptive Histogram Equalization) method in combination with color filtering, and the use of the DynamicCNN neural network denoiser for noise suppression. Experimental results showed that the ensemble approach using the soft voting strategy provides a statistically significant improvement in classification accuracy compared to individual models. The obtained results confirm the high efficiency of the proposed approach and demonstrate the potential of using ensemble deep learning models in medical diagnostics and clinical decision support tasks.

методы машинного обученияансамбль глубоких нейронных сетейрентгеновские снимкиденойзер DynamicCNNmachine learning methodsdeep neural network ensemblesX-ray imagesDynamicCNN denoiserРабота поддержана грантом Министерства науки и высшего образования Республики Казахстан (AP23488833).ReferencesLiu, X., Yu, Z., Tan, L. Deep learning for lung disease classification using transfer learning and a customized convolutional neural network architecture with attention. Proceedings of the IEEE 2nd International Conference on Sensors, Electronics and Computer Engineering (ICSECE), 2024, pp. 341–346.Liu, X., Yu, Z., Tan, L. Deep learning for lung disease classification using transfer learning and a customized convolutional neural network architecture with attention. Proceedings of the IEEE 2nd International Conference on Sensors, Electronics and Computer Engineering (ICSECE), 2024, pp. 341–346.Shamrat, F.M.J.M., Azam, S., Karim, A., Islam, R., Tasnim, Z., Ghosh, P., De Boer, F. LunGNeT22: A fine-tuned model for multiclass classification and prediction of lung disease using X-ray images. Journal of Personalized Medicine, 12, 680 (2022).Shamrat, F.M.J.M., Azam, S., Karim, A., Islam, R., Tasnim, Z., Ghosh, P., De Boer, F. LunGNeT22: A fine-tuned model for multiclass classification and prediction of lung disease using X-ray images. Journal of Personalized Medicine, 12, 680 (2022).De Sousa, P.M., Carneiro, P.C., Oliveira, M.M., Pereira, G.M., Da Costa, C.A. Jr., De Moura, L.V., Mattjie, C., Da Silva, A.M.M., Patrocinio, A.C. COVID-19 classification in X-ray chest images using a new convolutional neural network: CNN-COVID. Research on Biomedical Engineering, 38, 87–97 (2021).De Sousa, P.M., Carneiro, P.C., Oliveira, M.M., Pereira, G.M., Da Costa, C.A. Jr., De Moura, L.V., Mattjie, C., Da Silva, A.M.M., Patrocinio, A.C. COVID-19 classification in X-ray chest images using a new convolutional neural network: CNN-COVID. Research on Biomedical Engineering, 38, 87–97 (2021).Ashraf, S.M.N., Mamun, Md.A., Abdullah, H.Md., Alam, Md.G.R. SynthEnsemble: A fusion of convolutional neural networks, vision transformers, and hybrid models for multi-label chest X-ray classification. Proceedings of the 26th International Conference on Computer and Information Technology (ICCIT), 2023, pp. 1–6.Ashraf, S.M.N., Mamun, Md.A., Abdullah, H.Md., Alam, Md.G.R. SynthEnsemble: A fusion of convolutional neural networks, vision transformers, and hybrid models for multi-label chest X-ray classification. Proceedings of the 26th International Conference on Computer and Information Technology (ICCIT), 2023, pp. 1–6.Souid, A., Sakli, N., Sakli, H. Classification and prediction of lung diseases from chest X-ray images using MobileNetV2. Applied Sciences, 11, 2751 (2021).Souid, A., Sakli, N., Sakli, H. Classification and prediction of lung diseases from chest X-ray images using MobileNetV2. Applied Sciences, 11, 2751 (2021).Zhang, K., Zuo, W., Chen, Y., Meng, D., Zhang, L. Beyond a Gaussian denoiser: Residual learning of deep convolutional neural networks for image denoising. IEEE Transactions on Image Processing, 26, 3142–3155 (2017).Zhang, K., Zuo, W., Chen, Y., Meng, D., Zhang, L. Beyond a Gaussian denoiser: Residual learning of deep convolutional neural networks for image denoising. IEEE Transactions on Image Processing, 26, 3142–3155 (2017).Chest X-ray COVID-19 Pneumonia Dataset. Kaggle. URL: https://www.kaggle.com/datasets/tawsifurrahman/covid19-radiography-database (accessed: 05.01.2022).Chest X-ray COVID-19 Pneumonia Dataset. Kaggle. URL: https://www.kaggle.com/datasets/tawsifurrahman/covid19-radiography-database (accessed: 05.01.2022).COVID Chest X-ray Image Dataset for Research. Kaggle (accessed 05 January 2022). URL: https:// www.kaggle.com/datasets/pranavraikokte/covid19-image-datasetCOVID Chest X-ray Image Dataset for Research. Kaggle (accessed 05 January 2022). URL: https:// www.kaggle.com/datasets/pranavraikokte/covid19-image-datasetChest X-ray Pneumonia, COVID-19, Tuberculosis Dataset. Kaggle. URL: https://www.kaggle.com/datasets/amanullahasraf/chest-xray-pneumonia-covid19-tuberculosis (accessed: 05.01.2022).Chest X-ray Pneumonia, COVID-19, Tuberculosis Dataset. Kaggle. URL: https://www.kaggle.com/datasets/amanullahasraf/chest-xray-pneumonia-covid19-tuberculosis (accessed: 05.01.2022).COVID-19 Chest X-ray Dataset. GitHub. URL: https://github.com/ieee8023/covid-chestxray-dataset (accessed: 03.01.2022).COVID-19 Chest X-ray Dataset. GitHub. URL: https://github.com/ieee8023/covid-chestxray-dataset (accessed: 03.01.2022).COVIDx CXR-2 Dataset. Kaggle. URL: https://www.kaggle.com/datasets/andyczhao/covidx-cxr2 (accessed: 04.01. 2022).COVIDx CXR-2 Dataset. Kaggle. URL: https://www.kaggle.com/datasets/andyczhao/covidx-cxr2 (accessed: 04.01. 2022).Tuberculosis Chest X-ray Dataset. Kaggle. URL: https://www.kaggle.com/datasets/tawsifurrahman/tuberculosis-tb-chest-xray-dataset (accessed: 04.01.2022).Tuberculosis Chest X-ray Dataset. Kaggle. URL: https://www.kaggle.com/datasets/tawsifurrahman/tuberculosis-tb-chest-xray-dataset (accessed: 04.01.2022).Tuberculosis Chest X-rays Shenzhen Dataset. Kaggle. URL: https://www.kaggle.com/datasets/raddar/tuberculosis-chest-xrays-shenzhen (accessed: 05.01.2022).Tuberculosis Chest X-rays Shenzhen Dataset. Kaggle. URL: https://www.kaggle.com/datasets/raddar/tuberculosis-chest-xrays-shenzhen (accessed: 05.01.2022).COVID-19 Detection Dataset. Kaggle. URL: https://www.kaggle.com/datasets/darshan1504/covid19-detection-xray-dataset (accessed: 05.01.2022).COVID-19 Detection Dataset. Kaggle. URL: https://www.kaggle.com/datasets/darshan1504/covid19-detection-xray-dataset (accessed: 05.01.2022).Pneumothorax Binary Classification Dataset. Kaggle. URL: https://www.kaggle.com/datasets/vbookshelf/pneumothorax-chest-xray-images (accessed: 05.01.2022).Pneumothorax Binary Classification Dataset. Kaggle. URL: https://www.kaggle.com/datasets/vbookshelf/pneumothorax-chest-xray-images (accessed: 05.01.2022).National Institutes of Health Chest X-ray Dataset (Sample). Kaggle. URL: https://www.kaggle.com/datasets/nih-chest-xrays/sample (accessed: 06.01.2022).National Institutes of Health Chest X-ray Dataset (Sample). Kaggle. URL: https://www.kaggle.com/datasets/nih-chest-xrays/sample (accessed: 06.01.2022).COVID Normal Viral Opacity Dataset Version 2. Kaggle. URL: https://www.kaggle.com/datasets/tawsifurrahman/covid19-radiography-database (accessed: 03.01.2022).COVID Normal Viral Opacity Dataset Version 2. Kaggle. URL: https://www.kaggle.com/datasets/tawsifurrahman/covid19-radiography-database (accessed: 03.01.2022).X-ray Report Dataset. Kaggle. URL: https://www.kaggle.com/datasets/raddar/xray-report-dataset (accessed: 04.01.2022).X-ray Report Dataset. Kaggle. URL: https://www.kaggle.com/datasets/raddar/xray-report-dataset (accessed: 04.01.2022).ChestX-ray14 Dataset. National Institutes of Health. URL: https://nihcc.app.box.com/v/ChestXray-NIHCC (accessed: 03.01.022).ChestX-ray14 Dataset. National Institutes of Health. URL: https://nihcc.app.box.com/v/ChestXray-NIHCC (accessed: 03.01.022).Worldwide Chest X-ray Datasets. Kaggle. URL: https://www.kaggle.com/datasets/nikhilpandey360/chest-xray-multiple-datasets (accessed: 03.01.022).Worldwide Chest X-ray Datasets. Kaggle. URL: https://www.kaggle.com/datasets/nikhilpandey360/chest-xray-multiple-datasets (accessed: 03.01.022).COVID-19 Radiography Database. Kaggle. URL: https://www.kaggle.com/datasets/tawsifurrahman/covid19-radiography-database (accessed: 06.01.2022).COVID-19 Radiography Database. Kaggle. URL: https://www.kaggle.com/datasets/tawsifurrahman/covid19-radiography-database (accessed: 06.01.2022).Tuberculosis Chest X-rays Shenzhen Dataset. Kaggle. URL: https://www.kaggle.com/datasets/raddar/tuberculosis-chest-xrays-shenzhen (accessed: 06.01.2022).Tuberculosis Chest X-rays Shenzhen Dataset. Kaggle. URL: https://www.kaggle.com/datasets/raddar/tuberculosis-chest-xrays-shenzhen (accessed: 06.01.2022).Hussein, F., Mughaid, A., AlZu’bi, S., El-Salhi, S.M., Abuhaija, B., Abualigah, L., Gandomi, A.H. Hybrid contrast limited adaptive histogram equalization and convolutional neural network deep models for lung disease classification from X-ray images. Electronics, 11, 3075 (2022).Hussein, F., Mughaid, A., AlZu’bi, S., El-Salhi, S.M., Abuhaija, B., Abualigah, L., Gandomi, A.H. Hybrid contrast limited adaptive histogram equalization and convolutional neural network deep models for lung disease classification from X-ray images. Electronics, 11, 3075 (2022).Sonali, N., Sahu, S., Singh, A.K., Ghrera, S.P., Elhoseny, M. An approach for denoising and contrast enhancement of retinal fundus images using contrast limited adaptive histogram equalization. Optics and Laser Technology, 110, 87–98 (2018).Sonali, N., Sahu, S., Singh, A.K., Ghrera, S.P., Elhoseny, M. An approach for denoising and contrast enhancement of retinal fundus images using contrast limited adaptive histogram equalization. Optics and Laser Technology, 110, 87–98 (2018).Das, H., Rout, J.K., Moharana, S.C., Dey, N. Applied intelligent decision making in machine learning. CRC Press, Boca Raton, 320 p. (2020).Das, H., Rout, J.K., Moharana, S.C., Dey, N. Applied intelligent decision making in machine learning. CRC Press, Boca Raton, 320 p. (2020).Huang, G., Liu, Z., Van Der Maaten, L., Weinberger, K.Q. Densely connected convolutional networks. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017, pp. 4700–4708.Huang, G., Liu, Z., Van Der Maaten, L., Weinberger, K.Q. Densely connected convolutional networks. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017, pp. 4700–4708.Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., Chen, L.-C. MobileNetV2: Inverted residuals and linear bottlenecks. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018, pp. 4510–4520Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., Chen, L.-C. MobileNetV2: Inverted residuals and linear bottlenecks. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018, pp. 4510–4520Tan, M., Le, Q.V. EfficientNet: Rethinking model scaling for convolutional neural networks. arXiv preprint, arXiv:1905.11946 (2019).Tan, M., Le, Q.V. EfficientNet: Rethinking model scaling for convolutional neural networks. arXiv preprint, arXiv:1905.11946 (2019).Hu, J., Shen, L., Albanie, S., Sun, G., Wu, E. Squeeze-and-excitation networks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 42, 2011–2023 (2019).Hu, J., Shen, L., Albanie, S., Sun, G., Wu, E. Squeeze-and-excitation networks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 42, 2011–2023 (2019).Ma, N., Zhang, X., Zheng, H.-T., Sun, J. ShuffleNet V2: Practical guidelines for efficient convolutional neural network architecture design. Lecture Notes in Computer Science, 11218, 122–138 (2018).Ma, N., Zhang, X., Zheng, H.-T., Sun, J. ShuffleNet V2: Practical guidelines for efficient convolutional neural network architecture design. Lecture Notes in Computer Science, 11218, 122–138 (2018).Kingma, D.P., Ba, J.L. Adam: A method for stochastic optimization. arXiv preprint, arXiv:1412.6980 (2014).Kingma, D.P., Ba, J.L. Adam: A method for stochastic optimization. arXiv preprint, arXiv:1412.6980 (2014).

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