Moderní metody pro analýzu periostálního svalku z RTG obrazů pro sledování dynamiky procesu hojení kostí

Abstract

This thesis deals with the design and software implementation of segmentation methods that have the potential to automatically identify periosteal muscle from radiographic images in the context of the dynamics of healing bones. The segmentation method "active contours" was chosen to detect the periosteal muscle. The active contour method is a type of segmentation technique that can be defined as the use of energy forces and constraints to extract image points of interest for further processing and analysis. In the active contour method, multiple versions, called models, are available. The practical part of the work includes the development of a segmentation algorithm to segment the periosteal muscle. To develop the algorithm I have chosen Snake model which is suitable for medical image analysis. The individual segmentations are then objectively evaluated based onthe selected metrics against defined gold standards. In the subsequent part of the paper, selected muscle symptoms are extracted from the segmented images. The selected features are the perimeter, area, length and width, which form geometric parameters. Other features evaluated are the luminance parameters. These include the lowest, highest and average brightness values. These are used to assess fracture dynamics over time. The evaluated results show that when the difference in contrast between the periosteal muscle and bone region is large, the segmentation is very accurate because of this. As the fracture heals, the contrast decreases. This makes segmentation less accurate.