DEVELOPMENT OF A SOFTWARE ALGORITHM FOR OBTAINING MRI IMAGES OF THE BRAIN 

Evgeniy V. Bogdanov, Postgraduate student, Bauman Moscow State Technical University (building 1, 5 2nd Baumanskaya street, Moscow, Russia), E-mail: evgeniy.bogdanov95@gmail.com 

Background. Research into the formation of algorithms for obtaining MRI images is highly relevant. The purpose of the work is to analyze advances in the development of a software algorithm for obtaining MRI images of the brain. Materials and methods. Found 17 articles in specialized databases Cyberleninka, eLibrary, PubMed, ScienceDirect. Methods of analysis, synthesis and induction were used. Results. The best way to suppress noise in brain MRI images is the Gaussian Filter, the improvement of which has been achieved through the evolution of neural networks. Automatic segmentation achieved performance comparable to manual segmentation by using a combined system with integrated modules to eliminate the influence of noise and background, to identify image features and edge information. The Sobel operator allows the bright edges of an MRI image to be more clearly identified for removal. For volumetric visualization of brain images, due to its labor-intensive nature, the use of layer-by-layer representation of MRI data is proposed. Watershed segmentation and the K-nearest neighbor classification algorithm resulted in an MRI image accuracy of only 89 %; the wavelet transform was performed without calculating the accuracy. Support Vector Machine (SVM) using the GLCM algorithm showed an accuracy of up to 93 %, but only 36 images were used for training. Based on 150 MRI images of the brain, their classification was performed using the MATLAB 2018a software package (Matrix Laboratory) and a testing accuracy of 96,7 % was achieved. Conclusions. Improved algorithms for removing noise and bright edges from MRI brain images, segmenting them, and creating them with volumetric visualization are being created, including effective software modules for automatic segmentation based on convolutional neural networks. 

MRI, images, algorithm, noise removal, segmentation, filter, edges 

Bogdanov E.V. Development of a software algorithm for obtaining MRI images of the brain. Izmerenie. Monitoring. Upravlenie. Kontrol' = Measuring. Monitoring. Management. Control. 2024;(3):111–118. (In Russ.). doi: 10.21685/2307-5538-2024-3-13