G.S. Baydin – Lecturer, Bauman Moscow State Technical University (National Research University) 

E-mail: baydin1015@gmail.com

D.D. Bukreev – Student, Bauman Moscow State Technical University (National Research University)

E-mail: bukreevdanil09@gmail.com

The problem of the appearance of malignant neoplasms in the human stomach in the images of computed tomography is considered. With the increasing complexity of diagnosing malignant neoplasms in the human stomach, storing digital research results and increasing their volume, it becomes necessary to automate the clinical diagnosis of this disease, which can improve the accuracy and reliability of its results. Currently, doctors and researchers currently have enough medical records. Manual analysis of sufficiently large volumes of these images is a time-consuming task.

The article proposes one of the options for solving problems of automated recognition of malignant neoplasms in the human stomach in the images of computed tomography using image processing algorithms. As part of this solution, analysis and comparison of modern algorithms has been carried out. When analyzing images of computed tomography of the human stomach, it was found that their preliminary processing is required to increase the likelihood of success in recognizing malignant neoplasms. It is required to sharpen the contours of objects. An additional analysis of various algorithms for improving images was carried out, the results of which led to the conclusion that the solution to the problem requires the consistent application of the following image processing algorithms (filters): the first algorithm uses “blur” using a low-pass filter, the second algorithm uses image conversion in monochrome format.

The following image processing algorithms were considered to highlight the stomach region in the image: the GrabCut segmentation algorithm, the Mean Shift clustering algorithm, and the Watershed segmentation algorithm. A brief description of each of the listed algorithms and their comparative analysis in the framework of solving the problem, based on qualitative and quantitative metrics, based on which the GrabCut algorithm is selected, is given.

The developed technique on the tests samples of images of computed tomography images of the human stomach with and without malignant neoplasms (healthy people) in the automated mode correctly determined the presence of malignant neoplasms in the human stomach and its contours in 98% of cases. According to the results of the research, the optimal sequence of applying various algorithms for processing computer tomography images to solve the problem is obtained. The practical significance of the proposed methodology is its use at various stages of the diagnosis of malignant neoplasms in the human stomach in computed tomography images.

1. Hatakeyama M., Higashi H. Helicobacter pylori CagA: a new paradigm for bacterial carcinogenesis // Cancer Science. 2005. V. 96 (12). P. 835–843. doi: 10.1111/j.1349-7006.2005.00130.x.
2. Parkin D.M., Bray F., Ferlay J., Pisani P. Global cancer statistics, 2002. CA Cancer J. Clin. 2005. P. 74–108.
3. Berdichevskij E.G. Primenenie teorii raspoznavaniya obrazov k komp'yuternoj rekonstrukcii fresok // Fiziko-matematicheskie nauki i informacionnye tekhnologii: problemy i tendencii razvitiya: Sb. st. po mater. I Mezhdunar. nauch.-prakt. konf. Novosibirsk: SibAK. 2011. 565 s.
4. Zolotyh N.YU., Kozinov E.A., Kustikova V.D., Meerov I.B., Polovinkin A.N. Ob odnom metode povysheniya skorosti poiska ob"ektov metodom skrytyh opornyh vektorov za schet primeneniya kaskadnyh skhem // 11-ya Vseros. konf. Vysokoproizvoditel'nye parallel'nye vychisleniya na klasternyh sistemah (HPC–2011): Trudy konferencii. N. Novgorod: Izd. NNGU. 2011. C. 134–139.
5. Blum M., Springenberg J.T., Wlfing J., Riedmiller M. A Learned Feature Descriptor for Object Recognition in RGB-D Data // In: Proceedings of the IEEE International Conference on Robotics and Automation (ICRA). St. Paul, Minnesota. USA. 2012. P. 1–13.
6. Taleb A.A., Doubois P., Duquenoy E. Analysis Methods of CT-scan Images for the Characterization of the Bone Texture. Pattern Recognition 24. 2003. P. 35–48.
7. Pillow (PIL Fork) 3.1.2. documentation [Elektronnyj resurs]. Rezhim dostupa: https://pillow.readthedocs.io/en/3.1.x/reference/Image.html (Data obrashcheniya: 13.06.2019).
8. Ven Rajzina Dzh. Klassifikaciya i klaster. M.: Mir. 1980. 390 s.
9. Rother S., Kolmogorov V., Blake A. Grabcut – interactive foreground extraction using iterated graph cuts. 2004. P. 18–19.
10. Comaniciu D., Ramesh V., Meer P. Real-Time Tracking of Non-Rigid Objects Using Mean Shift, Conference on CVPR. 2000. V. 2. P. 1–8.
11. Najman L., Schmitt M. Watershed of a continuous function // In Signal Processing (Special issue on Mathematical Morphology.). V. 38. 1994. P. 99–112.
12. Koshunin V., Vezhnevec V. Metody segmentacii izobrazhenij: interaktivnaya segmentaciya // Komp'yuternaya grafika i mul'timedia. 2007. № 5. C. 13–18.
13. Kim E., Wang W., Li H., Huang X. A parallel annealing methods for automatic color cervigram image segmentation. Medical Image Computing and Computer Assisted Intervention. MICCAI-GRID. 2009. P. 88–91.
14. Nacional'naya biblioteka im. N.E. Baumana [Elektronnyj resurs]. CHichvarin N.V. Metod vodorazdelov (segmentaciya). Rezhim dostupa: https://ru.bmstu.wiki/Metod_vodorazdelov_(Segmentaciya) (Data obrashcheniya: 13.06.2019).
15. Roerdink Jos B.T.M., Meijster A. The Watershed Transform: Definitions, Algorithms and Parallelization Strategies // Fundamenta Informaticae. 2000. № 41. P. 187–228.
16. Szekely G.J., Rizzo M.L., Bakirov N.K. Measuring and testing independence by correlation of distances. Annals of Statistics. 2007. V. 35(6). P. 2769– 2794. doi: 10.1214/009053607000000505.
17. Yang W., Xu L., Chen X., Zheng F., Liu Y. Chi-Squared Distance Metric Learning for Histogram Data. Mathematical Problems in Engineering. 2015, Article ID 352849, 2015. P. 18–22.
18. OpenCV Open Source Computer Vision [Elektronnyj resurs]. Rezhim dostupa: https://docs.opencv.org/3.4/d8/dc8/tutorial_histogram_comparison.html (Data obrashcheniya: 12.06.2019).
19. Titov I.O., Emel'yanov G.M. Vydelenie konturov dvizhushchegosya ob"ekta // Vestnik Novgorodskogo gosudarstvennogo universiteta. 2010. № 55. S. 27–31.
20. Ramadevi Y., Sridevi T., Poornima B., Kalyani B. Image segmentation and object recognition based on edge detection technologies // International Journal of Computer Science and Information Technology (IJCSIT). 2010. V. 2. № 6. P. 35–42.
21. Nicolescu Cr., Alberts B., Jonker P. Parallel watershed algorithm on 2d images from cranial CT-scans using PVM and MPI on a distributed machine // Delf. University of Technology Faculty of Applied Physics, Pattern Recognition Group. Netherlands. 2015. V. 1. P. 34–38.
22. Chan T., Vese L. Active contours without edges // IEEE Transactions on Image Processing. 2001. V. 10. P. 266–277.
23. OpenCV Open Source Computer Vision [Elektronnyj resurs]. Rezhim dostupa: https://docs.opencv.org/3.4/da/d5c/tutorial_canny_detector.html (Data obrashcheniya: 12.06.2019).