The work is devoted to solving the actual scientific problem of developing and improving information decisions for medical image processing. The manuscript is consisted of the introduction, four sections, the conclusion, the references list and eight annexes.
In the first section «Modern information and communication solutions in e-health», the Ukrainian medical information systems were analyzed, their advantages and disadvantages were defined. The relevance of developing information technology of the archival medical data processing in particular images was shown. The information technology of the archival medical data processing which is based providing the continuous lifelong medical care was proposed. The proposed information technology includes the processes of preprocessing, classifying, automated archiving medical images and signals, protection data, as well as segmentation and features extraction of the images, description and retrieval of the objects on the images.
In the second section «Medical image classification», main approaches to image classification were analyzed. The gradational existing methods of the image processing for the brightness and contrast enhancement of the images were considered. The relevance of developing classification methods for archival medical images and also methods of the brightness and contrast enhancement of medical images was shown.
The classification method of digital and digitized medical images was developed. The parametric space is determined by metadata of images and also qualitative and quantitative features. The method of the image brightness enhancement which is based on the applying nonlinear transformations to the color components of the images for the HSV color model was improved. The application of the proposed method allows to increase the accuracy of contour detection on 31.83±5.81% compared to the method of linear stretching.
In the third section «The medical image segmentation», the segmentation methods were analyzed, their advantages and disadvantages were defined. The relevance of the medical image transformation methods development was shown. For the first time the automated segmentation method of the homogeneous medical images was developed. The proposed segmentation method of the medical images is based on the analysis of the statistical indicators of the input image set. The results of the applying proposed method is the image set of the background, the component parts of the initial images and the contours.
In the fourth section «Image vectorization and retrieval», the methods of the image contour detection were analyzed, their advantages and disadvantages were defined. The methods of the medical image retrieval were analyzed, their advantages and disadvantages were defined. The relevance of developing the methods of the image processing for the vector analysis of the medical images was shown.
For the first time the contour description method of the surveillance objects on the video images and the static medical images was developed. The proposed method is based on the contour representation of the images and its objects by the coordinate sets of «reference» vertices, the size of which is input parameter of the method and can change in depending on the requirements, which are defined by the tasks of the retrieval and required reproduction accuracy. The result of applying method is vector surveillance objects on the images, which are described by the continuous contour curves. Compared to the existing methods of the interpolation (Differential Equation Interpolant), the application of the proposed method allows to decrease the volume of the graphical data which are saved and processed in medical software and hardware systems in 2.71±0.53 times. For the first time the fuzzy retrieval method of the objects on medical images which is based on the definition of the range of the permissible values of the vertex coordinates for vector images to estimate the similarity degree of the geometric shapes for the candidate-images to the template-images was developed. The applying two stage estimation of the similarity degree of images allows to quantifiable define the following categories of candidate-images according to template-images: «identical», «similar», «slightly similar», and «dissimilar». The application of the proposed method allows to increase the retrieval accuracy in 20.23±2.14% compared to the content-based image retrieval method. The smoothing method of the contour curves of the image objects was improved. The proposed method is based on the applying the nonlinear operations of the coordinate transformations to the contour vertices and allows to suppress the noises on the medical images represented in the vector format. The application of the proposed method allows to increase the reproduction accuracy of the image objects in 10.96±1.53% compared to existing median filtering method.
