Improving the quality and efficiency of identification of special states of monitored objects based on the development of mathematical and software for processing computer images using large databases

The relevance of the study is due to an increase in human diseases, which are associated with significant socio-economic damage and give a significant burden on health. According to WHO recommendations, a disease prevention system should include prevalence assessment, correction, and risk factor management (WHO, 2009). A special place in this set of measures is occupied by the mass disease monitoring system, both a mechanism for assessing the situation and the need for implementing preventive measures, and a method for monitoring the effectiveness of implemented preventive measures. In this regard, this article considers the creation of an algorithm for processing images of a computer tomography scan of a human lung using software. The leading method to study this problem are neural networks. The article presents a convolutional neural network model of Chexnet X-ray processing developed by scientists from Stanford University. An algorithm for developing a mechanism for analyzing images based on modern x-ray images of organs - computed tomography images, which are obtained using a complex software and hardware complex with ultra-sensitive detectors for recording x-ray radiation, as well as an extensive software package that allows you to obtain images with high spatial resolution, is considered. The developed algorithm is implemented on the basis of the Densenet convolution network, the depth of which is 201 layers. Changes were made to it in the form of using the ReLU activation function (short for English rectified linear unit), which can significantly speed up the learning process and at the same time significantly simplify calculations. As a result, the developed convolutional neural network helps the continuity of data collection, which allows to improve the process of strategic decision-making, to develop action programs in the field of public health

1. Belikova T.P. Modelirovaniye lineynykh fil'trov dlya obrabotki rentgenovskikh izobrazheniy v zadachakh meditsinskoy diagnostiki. Tsifrovaya optika. Obrabotka izobrazheniy i poley v eksperimental'nykh issledovaniyakh. M.: Nauka, 1990. (In Russ)

2. Gonsales R. Tsifrovaya obrabotka izobrazheniy: per. s angl. R. Gonsales, R. Vuds. – M.: Tekhnosfera, 2005. (In Russ)

3. Trufanov G.Ye., Ryazanov V.V., Dergunova N.I. i dr. Sovmeshchennaya pozitronnoemissonnaya i komp'yuternaya tomografiya (PET-KT) v onkologii. Elbi–sPb, 2005. (In Russ)

4. Deng, Jia, Dοng, Wei, Sοᴄher, Riᴄhard, Li, Li–Jia, Li, Kai, and Fei–Fei, Li. Imagenet: A large–sᴄale hierarᴄhiᴄal image database. In Cοmputer Visiοn and Pattern Reᴄοgnitiοn, 2009;248–255.

5. Mistyukov B.V., Burkovskiy V.L Neyrosetevoye modelirovaniye dinamiki nelineynykh sistem. Vestnik voronezhskogo gosudarstvennogo tekhnicheskogo universiteta, 2012;3(12- 1):51-56. (In Russ)

6. Vasil'chenko V.A., Burkovskiy V.L. Analiz i vybor struktury ob"yektno-relyatsionnoy SUBD meditsinskoy informatsionnoy sistemy. Vestnik voronezhskogo gosudarstvennogo tekhnicheskogo universiteta, 2015;3(3):57–59. (In Russ)

7. Vasilchenko V.A., Burkovskiy V.L. Struktura ekspertnoy sistemy operativnoy diagnostiki i lecheniya zabolevaniy legkikh. Vestnik voronezhskogo gosudarstvennogo tekhnicheskogo universiteta, 2016;12 (1):28-31. (In Russ)

8. Goncharova YU.M., P'yanikov A.Ye, Burkovskiy V.L. Neyrosetevoye modelirovaniye ostorogo piyelonifrita v eksperimente. Vestnik voronezhskogo gosudarstvennogo tekhnicheskogo universiteta. 2012;8(7-1):54-57. (In Russ)

9. Vasilchenko VA, Burkovskiy VL, Danilov AD. Algorithmization of the process of recognition of states of living objects based on special x-ray images. Computer Optics 2019; 43(2): 296-303. DOI: 10.18287/2412-6179-2017-43-2-296-303.