APPLICATION OF MACHINE LEARNING METHODS OF NEURAL NETWORKS TO SOLVE THE PROBLEM OF SKIN DISEASES DIAGNOSING
DOI:
https://doi.org/10.32782/IT/2022-3-2Keywords:
machine learning, diagnosis, neural network, clustering, data, training, processing, imageAbstract
The article is devoted to the study of the application of machine learning technology for skin disease diagnosis. The object of the study is the process of recognizing and classifying skin diseases based on their photographs. The relevance of the research is due to the fact that today the methods of artificial intelligence are widely used in medicine and allow diagnosing diseases in those cases when an in-person visit to the doctor is difficult for certain reasons, in particular, in telemedicine. The purpose of the work is to develop the neural network model for diagnosing skin diseases and its further implementation in an information system, which will be able to recognize and classify diseases by the provided photo of the skin area. The paper presents the application of the image processing method, namely the Kmeans clustering algorithm for improving the quality of images in the input dataset. The convolutional neural network CNN to classify skin diseases was used. An input dataset the DermNet dataset with images was used, which was previously divided into a training and a test sample. Data processing, namely the selection of the localization of the disease on the image, was performed using mathematical calculations, namely the Sklearn library. After processing, the data is sent to a pretrained convolutional neural network, which was built using the TensorFlow framework. According to the results of neural network training, where all calculations were not performed on a graphics processor, the recognition accuracy of the entire test sample was more than 0.7, and for certain images, it reached 0.9. This result was obtained under the condition that the data should be put into the clustering algorithm before training, or in other words, the initial processing should be carried out. Thus, as a result of the conducted research, it can be concluded that the K-means method is suitable for solving problems of image preprocessing in dermatology, and there is no conflict in the process of training a model based on the TensorFlow framework with standard network parameters. The further research direction is to improve the classification accuracy of the developed neural network by optimizing the parameters on different neural network architectures, as well as using ensemble methods to investigate the improvement and speed of the neural network.
References
Alkolifi Alenezi N. S. A Method of Skin Disease Detection Using Image Processing and Machine Learning. Procedia Computer Science. 2019. Vol. 163. URL: https://doi.org/10.1016/j.procs.2019.12.090 (date of access: 27.02.2023).
Improve image registration jeffrey’s divergence method for insufficient overlap area using kmeans++ in remote sensed images / Alnagdawi M. A. et al. Journal of Theoretical and Applied Information Technology. 2019. Vol. 97 (5). P. 1571-1580.
Machine Learning and Deep Learning Integration for Skin Diseases Prediction / Bandyopadhyay, S. K. et al. International Journal of Engineering Trends and Technology. 2022. Vol. 70. No. 3. P. 13-21 URL: https://doi.org/10.14445/22315381/IJETT-V70I2P202 (date of access: 27.02.2023).
Barakbah A. R., & Kiyoki Y. A new approach for image segmentation using Pillar-Kmeans algorithm. World Academy of Science, Engineering and Technology. 2009. Vol. 59.
Chauhan R., Ghanshala K. K., & Joshi R. C. Convolutional Neural Network (CNN) for Image Detection and Recognition. Proceedings of the ICSCCC 2018 - 1st International Conference on Secure Cyber Computing and Communications. 2018. URL: https://doi.org/10.1109/ICSCCC.2018.8703316 (date of access: 27.02.2023).
Research on the influence of kmeans cluster preprocessing on adversarial images / Chen G. et al. ACM International Conference Proceeding Series. 2019. P. 248-252. URL:https://doi.org/10.1145/3377170.3377220 (date of access: 27.02.2023).
Corn Leaf Image Segmentation Based on Improved Kmeans Algorithm / Guo, C. F. et al. Journal of North University of China (Natural Science Edition). 2021. Vol. 42 (6). URL:https://doi.org/10.3969/j.issn.1673-3193. 2021.06.007 (date of access: 27.02.2023).
Hajesmaeel-Gohari S., & Bahaadinbeigy K. The most used questionnaires for evaluating telemedicine services. BMC Medical Informatics and Decision Making. 2021. Vol. 21 (1). URL: https://doi.org/10.1186/s12911-021-01407-y (date of access: 27.02.2023).
Classification of Rastrelliger kanagurta and Rastrelliger brachysoma using Convulational Neural Network (CNN) / Kurniawa K. et al. IOP Conference Series: Earth and Environmental Science. 2022. Vol. 969 (1). URL:https://doi.org/10.1088/1755-1315/969/1/012017 (date of access: 27.02.2023).
Automatic skin disease diagnosis using deep learning from clinical image and patient information / Muhaba K. A. et al. Skin Health and Disease. 2022. Vol. 2(1). URL:https://doi.org/10.1002/ski2.81 (date of access: 27.02.2023).
Telemedicine Practice: Review of the Current Ethical and Legal Challenges / Nittari G. et al. Telemedicine and e-Health. 2020. Vol. 26 (12). No. 1427-1437. URL:https://doi.org/10.1089/tmj.2019.0158 (date of access: 27.02.2023).
CNN-based person detection using infrared images for night-time intrusion warning systems. Park J. et al. Sensors (Switzerland). 2020. Vol. 20(1). URL:https://doi.org/10.3390/s20010034 (date of access: 27.02.2023).
Automated skin disease identification using deep learning algorithm / Patnaik S. K. et al. Biomedical and Pharmacology Journal. 2018. Vol. 11(3). URL: https://doi.org/10.13005/bpj/1507 (date of access: 27.02.2023)
Rajasekaran G., Aiswarya, N., & Keerthana, R. Skin Disease Identification Using Image Processing and Machine Learning Techniques. International Research Journal of Engineering and Technology. 2020. Vol. 7. Issue 3. P. 1368-1371.
AI-based localization and classification of skin disease with erythema /Son H. M. et al. Scientific Reports. 2021. Vol. 11(1). URL:https://doi.org/10.1038/s41598-021-84593-z (date of access: 27.02.2023).
Talpur, S., & Khoso, N. Advanced Ambulatory Operating Stretcher Learned by Means of Convulational Neural Network (CNN). Journal of Biomedical Engineering and Medical Imaging. 2018. Vol. 5 (3). URL:https://doi.org/10.14738/jbemi.53.4660 (date of access: 27.02.2023).
Allergic and non-allergic skin diseases in children of Ukraine: a retrospective study of the prevalence and incidence over the past 24 years /Volosovets O. P. et al. Medicni Perspektivi, 2021. Vol. 26 (3). URL:https://doi.org/10.26641/2307-0404.2021.3.242265 (date of access: 27.02.2023).
Gradient Color Leaf Image Segmentation Algorithm Based on Meanshift and Kmeans /Wang Q. et al. Proceedings of the IEEE Advanced Information Technology, Electronic and Automation Control Conference (IAEAC). 2021. URL: https://doi.org/10.1109/IAEAC50856.2021.9391029 (date of access: 27.02.2023).
Application of deep convulational neural network in medical image classification / Yadav S. et at. Proceedings of the 2021 International Conference on Emerging Smart Computing and Informatics. 2021. https://doi.org/10.1109/ESCI50559.2021.9396854 (date of access: 27.02.2023).
