Features of constructing solutions of a genetic algorithm in the problem of recognizing patterns of pneumonia disease

• Main Authors: Fedorchenko, Ye. M., Oliinyk, A. O., Korniienko, S. K., Kharchenko, A. S., Goncharenko, D. A.
• Format: Article
• Language: Ukrainian
• Published: Інститут проблем реєстрації інформації НАН України 2020
• Subjects: генетичний алгоритм; еволюційний алгоритм; рентгенограма; розпізнавання зображень; neural network; long short — term memory; genetic algorithm; evolutional algorithm; x-ray; representative image; long short-term memory
• Online Access: http://drsp.ipri.kiev.ua/article/view/218978

Modern medicine depends on advances in medical instrumentation and medical software development. One of the most important tasks facing doctors is to determine the exact boundaries of tumors and other abnormal formations in the tissues of the human body. The problems and methods of machine classification and recognition of images in the form of radiographs, as well as the issues of improving artificial neural networks, which are used to improve the quality and accuracy of detecting abnormal structures on chest radiographs, are discussed. A modified genetic method for optimizing model parameters based on convolutional neural networks has been developed to solve the problem of recognizing diagnostically significant signs of pneumonia on an X-ray image of the lungs. The fundamental difference between the proposed genetic method and existing analogs lies in the use of a special mutation operator in the form of an additive convolution of two mutation operators, which allows to reduce the training time of the neural network, as well as to identify the most suitable for studying the «margin of solutions». A comparative assessment of the effectiveness of the proposed method and known methods was presented, which showed an improvement in accuracy when solving the problem of searching for signs of pathology on an x-ray of the lungs. The practical use of the developed method will reduce labor intensity, increase the reliability of the search, speed up the process of diagnosing diseases and reduce the number of errors and repeated examinations of patients. Tabl.: 5. Fig.: 10. Refs: 34 titles