Defense of the dissertation of Kassylkassova Kamila for the degree of Doctor of Philosophy (PhD) in the specialty «8D06104 - Computer Engineering and Software»
L.N. Gumilyov Eurasian National University, a dissertation defense for the degree of Doctor of Philosophy (PhD) by Kassylkassova Kamila on the topic «Development of algorithms and software for solving of healthcare problems» to the educational program «8D06104 - Computer Engineering and Software».
The dissertation was carried out at the «Сomputer and software engineering education department» of L.N. Gumilyov Eurasian National University.
The language of defense is russian
Official reviewers:
Cherikbaeva Lyailya Sharipovna - Doctor of Philosophy (PhD), Associate Professor, Department of Computer Science, al-Farabi Kazakh national university (Almaty, Republic of Kazakhstan).
Ismailova Aisulu Abzhapparovna - Doctor of Philosophy (PhD), Associate Professor, Department of Information Systems, S.Seifullin Kazakh agrotechnical research university (Astana, Republic of Kazakhstan).
Temporary members of the Dissertation Committee:
- Vladimir Borisovich Barakhnin - Doctor of Technical Sciences, Professor, Head of the Department of Mathematical Modeling, Faculty of Mechanics and Mathematics, Novosibirsk State University (Novosibirsk, Russian Federation).
- Timur Zhumakanovich Merembaev - Doctor of Philosophy (PhD), Senior Researcher, RSE “Institute of Information and Computational Technologies” of the KN MNVO RK (Almaty, Republic of Kazakhstan).
- Saya Zamanbekovna Sapakova - Candidate of Physical and Mathematical Sciences, Assistant Professor, International University of Information Technologies (Almaty, Republic of Kazakhstan).
Scientific advisors:
- Yessengalieva Zhanna Serzhanovna - Doctor of Philosophy (PhD), Department of Computer and Software Engineering of the L.N. Gumilyov Eurasian national university.
- Urazboev Gayrat Urazboevich - Doctor of Physical and Mathematical Sciences, Professor of the Urgench State University (Urgench, Uzbekistan).
The defense will take place on August 12, 2025, at 02:00 PM in the Dissertation Council for the training direction «8D061 - Information and communication technologies» in the specialty «8D06104 - Computer Engineering and Software» of L.N. Gumilyov Eurasian National University. The defense meeting is planned to be held online.
Link: https://clck.ru/3MyJEj
Address: Astana, A. Pushkin str., 11, educational building, auditorium 222.
Abstract (English): ANNOTATION dissertation work of Kassylkassova Kamila «Development of algorithms and software for solving healthcare problems», submitted for the degree of Doctor of Philosophy (PhD) in the education program «8D06104 - Computer Science and Software» Relevance of the research topic. Since the outbreak of COVID-19 in December 2019, no one foresaw the scale of this disease. As of August 09, 2023, there were more than 760 million cases of infection worldwide and more than 6.9 million deaths. The pandemic has shown that medicine around the world is not ready to cope with such a huge number of patients. To mitigate the impact of pandemics in the future on healthcare, where one of the popular ways is to introduce medical software into the daily life of citizens using innovative technologies such as integrated mobile healthcare applications, Bluetooth, GPS, artificial intelligence and machine learning, where technologies can significantly improve the remote provision of medical services while observing preventive measures, namely social distancing and home quarantine. In accordance with the strategy «Kazakhstan-2050», administrative bodies strive to ensure high quality and safety of medical care by standardizing all processes in medical institutions. An important aspect is the development and improvement of clinical protocols and standards of specialized services based on advanced technologies and achievements of medical science. Rural healthcare in Kazakhstan faces a number of challenges, including remoteness from central areas, insufficient infrastructure, limited resources, aggressive climatic conditions, shortage of medical personnel and high staff turnover. The problems are caused by the insufficient importance of primary healthcare in the field of prevention, as well as limited public awareness of the importance of a healthy lifestyle and disease prevention. At the primary level of healthcare, there is an insufficient development of general medical practice and the use of technologies that contribute to improving the overall health of the population, which requires additional development and improvement. In order to improve the quality of service and reduce queues, it is necessary to optimize management in primary healthcare organizations by introducing modern methods of operational management and queue management technologies. In the context of the current health of the population and the expected increase in non-communicable diseases, the introduction of an integrated model of organizing medical care is also relevant. The degree of scientific development of the topic. Pneumonia in the modern world remains one of the main causes of death, taking more than 2.5 million lives annually. Timely diagnosis of pneumonia using chest X-ray is an important tool. In addition to the advantages of X-ray availability for patients anywhere in the world, it should be taken into account that the symptoms of pneumonia on X-ray images are not always obvious to the doctor. In such cases, software can help in making an accurate diagnosis. In terms of modern pneumonia detection solutions, it is not surprising that computer vision is an important area of neural network research, primarily because it provides answers to a wide range of questions that people face today. Biomedical image analysis using neural networks is one area of computer vision that has been proven to be effective many times. In recent years, there has been an increase in the use of deep models, especially convolutional neural networks (CNNs), as the dominant method for categorizing clinical images. This is because the selection of features to extract in traditional computational intelligence approaches is a labor-intensive process that also varies depending on its object. These studies present CNNs with different architectural styles and methodological approaches that were conducted using X-ray images. In order to obtain more favorable results, CNN-based models require a significant number of training samples. Collecting medical images is quite challenging due to the process of identifying medical data, which is complicated by labor-intensive privacy regulations and explanations from healthcare professionals. According to the researchers, transform-based data augmentation has proven to be a suitable method for image classification. Image enhancement methods can help prevent overfitting during the training phase, which ultimately leads to a more accurate model. Most of the strategies discussed here use transfer learning, which means that the deep learning methods are initially trained on data unrelated to pneumonia diagnostics. Using convolutional neural networks built from scratch in a number of image processing algorithms has shown that a simpler structure can provide higher accuracy than the many pre-trained legacy models used in transfer learning. The PCAnet model, in which the parameters of the convolutional layer are initialized by extracting features from the principal components of the image, has shown good results in image recognition tasks. One of the papers reviewed combined a convolutional neural network and a recurrent neural network to propose a new deep learning structure. The CNN learns low-level features from the original image and uses them as input to the recurrent neural networks (RNN). The RNN then analyzes the high-level features, which allowed achieving high recognition accuracy in color-depth image processing tasks. A year later, this method was modified and a multi-scale convolutional recurrent neural network was proposed. Where we added local contrast normalization and sampling, which were used as input data for the RNN, thereby allowing us to extract more abstract high-level features. Although there are many CNN-based image recognition algorithms, their performance is highly dependent on the database used for training. Scientists continue to search for optimal parameters and algorithms to achieve the best results. However, the human factor plays a significant role in the training process, and to date there is no structured model that would fully explain the influence of the network structure on the recognition quality. Especially in the classification and recognition of natural images, the choice of initial network parameters and optimization algorithm has a significant impact on the training process. The aim of this research dissertation is to develop a method for processing medical images using neural networks and software implementation of some tasks in the field of healthcare. To achieve this goal, it is necessary to solve the following tasks: - scientific and technical analysis of modern medical software, where it is necessary to analyze the features, advantages and disadvantages; - collection and preparation of a database of chest X-ray images, including both positive and negative cases of pneumonia for training and testing the proposed neural network; - development of a method and architecture of a CNN capable of analyzing and classifying chest X-ray images for diagnosing pneumonia; - implementation of an algorithm and software for virtual medical consultations with automatic analysis of X-ray images for pneumonia. The object of the research is medical software with intelligent diagnostics of pneumonia and other chest diseases using convolutional and recurrent neural networks. The subjects of the research are models, methods and algorithms for processing medical images using artificial intelligence. Research methods. During the dissertation research, various methods were used, such as synthesis and analysis of works by foreign and domestic researchers, as well as modern approaches to the development of algorithms and software, which are based on the use of deep, recurrent and convolutional neural networks in healthcare tasks. The main provisions submitted for defense: 1. Architecture and method of convolutional and recurrent neural network using the idea of skipping the ResNet convolution layer, aimed at increasing the accuracy of diagnostics of respiratory diseases using radiographic images. 2. Medical software for remote medical care, combining modules for doctors and patients, which allows to increase the availability and quality of medical services provided. The theoretical and practical significance of the work lies in the development and improvement of algorithms based on the use of a deep CNN with 24 hidden layers for detecting pneumonia on chest X-rays. The practical value lies in the application of the developed algorithms and software for mobile medical care, improving the quality of diagnostics and monitoring of diseases. The results of the study have been implemented in Jysan Med LLP and in the work of the State Enterprise on the Right of Economic Management «Abai City Hospital». The research results were tested at publications, including international and national conferences: - Seminars of doctoral students of the Department of Computer and Software Engineering (Astana, 2022-2024); - Proceedings of the National Student Scientific Conference "Contribution of Youth Science to the Implementation of the Kazakhstan-2050 Strategy" (Karaganda, 2022); - Proceedings of the International Scientific and Practical Online Conference "Integration of Science, Education and Production - the Basis for the Implementation of the Nation's Plan" (Saginov Readings No. 13), dedicated to the 30th Anniversary of Independence of the Republic of Kazakhstan, June 17-18, 2021. - XX International Scientific Conference «ǴYLYM JÁNE BІLІM - 2025», April 11, 2025. - Eurasian International Scientific Conference «Artificial Intelligence and Inverse Problems in Science, Technology and Industry», April 14-16, 2025. Publications published based on the research results. Scientific journals indexed in Scopus: 1. Automated Pneumonia Diagnosis using a 2D Deep Convolutional Neural Network with Chest X-Ray Images // International Journal of Advanced Computer Science and Applications (IJACSA). - 2023. - Vol. 14, Issue 2. - P. 699-708. 2. Optimization method for integration of convolutional and recurrent neural network // Eurasian Journal of Mathematical and Computer Applications. - 2023. Vol. 11, Issue 2. - P. 40-56. Publications in journals recommended by the authorized body (SHEQAC MSHE RK): 1. Analysis of medical applications created specifically to combat COVID-19 // Bulletin of the NAS RK. Series Physics and Information Technology No. 1 (341), Almaty, 2022. - P. 34-42. 2. Data Processing in the Development of Healthcare Software in the Context of COVID-19 // Bulletin. Kazakh National Pedagogical University named after Abay Series «Physical and Mathematical Sciences» No. 1 (77), Almaty, 2022. - P. 99-105. 3. Comparative Analysis of SmartMed and Damumed Software // Republican Scientific and Technical Journal «University of Enbekteri - University Works» No. 2 (87), Karaganda, 2022. - P. 284-290. Publications in domestic scientific journals: 1. Development of software for healthcare in the context of COVID-19 // L.N. Gumilyov published his works on the topic of technology and technology. - 2021.- No. 1 (134). - P. 91-99. Certificate of entering information into the state register of rights to objects protected by copyright (Appendix B): 1. Certificate of entering information into the state register of rights to objects protected by copyright No. 36007 dated May 18, 2023. Software «SmartMed». The structure of the dissertation consists of the content, definitions and abbreviations, introduction, three sections, conclusion, list of references, appendices. The introduction reveals the relevance and formulates the problems of the research topic. The object and subject of the study are defined, the main idea of the work is described, the purpose and objectives of the study are formulated, the scientific novelty and practical significance of the work are presented. The first section contains an overview of the methods of digital processing of medical images, an analysis of existing approaches to automated diagnostics of diseases from X-ray images, as well as a description of modern neural network technologies used in medicine. The chapter discusses the areas of software application in healthcare, including the analysis of medical mobile applications. Particular attention is paid to data processing issues in the development of software for medical institutions. An analysis of the prospects for the development of remote technologies in healthcare, including telemedicine and remote patient monitoring systems is also presented. The possibilities of using convolutional neural networks for the analysis of medical images and their application in diagnostic systems are considered. The second section formulates the requirements for the system being developed, conducts a comparative analysis of existing solutions in the field of medical diagnostics using machine learning, and justifies the choice of the model architecture. Also, algorithms for pre-processing of medical images are developed and implemented, a convolutional neural network is trained and tested, and the accuracy of the proposed method is assessed. The third section is devoted to the development and implementation of a software package that includes modules for automated pre-processing of medical images, segmentation of areas of interest, and classification of identified pathologies. This section presents algorithms for processing radiographic images using deep learning methods, as well as the integration of convolutional neural networks with additional modules for interpreting the results. Experimental studies aimed at assessing the accuracy of the developed system and comparing it with existing diagnostic methods are included. The applicant has developed an intelligent model for analyzing radiographic images using deep learning methods. A software package for automated disease recognition has been developed, which can be integrated into medical information systems. The conclusion presents the main results of the dissertation, formulates conclusions on the conducted research, and describes possible directions for further development of the developed algorithms and software solutions. The appendices contain acts on the implementation of research results, examples of the operation of the developed software, as well as certificates of registration of copyrights for the created algorithms. In general, the work is completed in printed form on 103 pages, using computer capabilities for emphasizing attention in the form of illustrations, diagrams and tables. The list of references consists of 89 titles. The author expresses deep heartfelt gratitude to his scientific adviser, Doctor of Philosophy (PhD), associate professor of the Department of Computer and Software Engineering Zhanna Yessengaliyeva and foreign adviser, Honorary Professor of Urgench State University, Doctor of Physical and Mathematical Sciences Gayrat Urazboev.
'/%3e%3cpath%20fill-rule='evenodd'%20clip-rule='evenodd'%20d='M17.5427%2012.684C17.1383%2011.7889%2016.7126%2011.771%2016.3278%2011.7553C16.0132%2011.7419%2015.6529%2011.7427%2015.2932%2011.7427C14.9333%2011.7427%2014.3483%2011.8774%2013.8537%2012.4152C13.3587%2012.9531%2011.9639%2014.2533%2011.9639%2016.8978C11.9639%2019.5426%2013.8987%2022.098%2014.1684%2022.4569C14.4383%2022.8153%2017.9033%2028.4156%2023.3907%2030.5701C27.9513%2032.3606%2028.8793%2032.0045%2029.8692%2031.9149C30.8591%2031.8253%2033.0633%2030.6149%2033.5133%2029.3597C33.9632%2028.1048%2033.9632%2027.0291%2033.8282%2026.8043C33.6932%2026.5803%2033.3333%2026.4459%2032.7933%2026.1771C32.2534%2025.9083%2029.5992%2024.6078%2029.1043%2024.4286C28.6093%2024.2494%2028.2494%2024.1598%2027.8894%2024.6979C27.5295%2025.2355%2026.4954%2026.4459%2026.1804%2026.8043C25.8655%2027.1635%2025.5505%2027.2083%2025.0105%2026.9395C24.4706%2026.6699%2022.7318%2026.1029%2020.6691%2024.2717C19.0641%2022.8472%2017.9806%2021.0877%2017.6656%2020.5496C17.3506%2020.012%2017.6318%2019.7209%2017.9027%2019.4529C18.1451%2019.2121%2018.4426%2018.8254%2018.7126%2018.5116C18.982%2018.1977%2019.072%2017.9737%2019.252%2017.6153C19.4319%2017.2564%2019.342%2016.9425%2019.207%2016.6737C19.072%2016.405%2018.0228%2013.7468%2017.5427%2012.684Z'%20fill='white'/%3e%3cdefs%3e%3clinearGradient%20id='paint0_linear_181_11959'%20x1='22.8525'%20y1='42.8177'%20x2='22.8525'%20y2='0.921701'%20gradientUnits='userSpaceOnUse'%3e%3cstop%20stop-color='%2320B038'/%3e%3cstop%20offset='1'%20stop-color='%2360D66A'/%3e%3c/linearGradient%3e%3c/defs%3e%3c/svg%3e)
