Thesis:

Spinal column and its segmentation٫ MRI ۳D imaging

Vertebral bone segmentation from magnetic resonance (MR) images is a challenging task. And it is a process whose purpose is to extract the elements that are supposed to be further analyzed. In this research, one of the main issues in imaging diagnosis, which is the precise determination of the desired area, has been addressed. Medical image processing makes it necessary to distinguish the test object from the rest of the less important details of the image. To choose the correct segmentation method, the structure of the texture and the type of the image should be considered.Here, selected segmentation methods implemented in Matlab and results obtained using clinical software are presented.Various evaluations in orthopedic diagnosis and radiological light from scoliosis, osteoporotic stenosis or vertebral fracture to bone metastases rely on non-invasive medical imaging such as magnetic resonance imaging.A fully three-dimensional and fully unsupervised cross-synthesis method is presented for the segmentation of scoliosis spines. Despite advanced systems for image analysis and medical diagnosis, perfect results for object segmentation cannot be obtained.This is due to similar pixel values describing different textures. Therefore, special attention should be paid to the accurate selection of segmentation parameters. It is also possible to have individual pixels with a different value from the range assigned to a given texture. The methods implemented in Matlab are often used in edge detection in images with low interference content due to the low complexity of the algorithm and low computational cost.Absence of other bony structures in the cross section or defects in the vertebrae. Roberts masks are used to detect lines of opposite direction, they are very sensitive to interference. Prewitta mask is used to detect edges in the form of corners in different fixed directions. The segmentation results with the growth zone method depend on the choice of threshold. While region-growing segmentation gives good results for uncomplicated images, it tends to produce small regions for complex images.

Keywords: 3D, spine, MIR images