Characterization of structural changes in spinal vertebrae based on perturbations to an adaptive model

• Autor(a) principal: Bernardo, Duarte António do Espírito Santo
• Data de Publicação: 2022
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10362/157130
• topic: DISH; X-ray; Image processing; segmentation; Morphological Active contours; Domínio/Área Científica::Engenharia e Tecnologia::Outras Engenharias e Tecnologias

Diffuse Idiopathic Skeletal Hyperostosis, or DISH, is a disease characterized by ossification of the entheses and the anterior longitudinal ligament. The diagnosis is made by visual analysis of an X-ray by a professional using the Resnick Criterion. The different experience among professionals and the fact that this criterion is only suitable in advanced stages of the disease make diagnosis difficult. Therefore, this work aims to contribute to the development of an auxiliary diagnostic tool for this disease. For this, a semi-automatic vertebral segmentation algorithm based on active morphological contours was proposed, comparing it with previous work and with segmentations made by experts on two radiographic images. Next, the corners of the vertebrae, where the disease manifests itself, were analyzed in order to characterize images with DISH. To accomplish this, it was assumed symmetry of the vertebrae and a Gaussian distribution of the histograms of those corners to analyze them and calculate two ratios: Left upper corner mean value / Right upper corner mean value (LS/RS) and Left lower corner mean value / Right lower corner mean value (LI/RI), in order to find a differentiating metric between vertebrae with pathology and those without. The results achieved by the algorithm were clearly superior to the previous work and similar to that of the experts. The analysis of pathologic vertebrae revealed a difference in the shift of the distributions of pathologic corners relative to non-pathologic ones, which is not seen in vertebrae without apparent pathology. Regarding the ratios, the LI/RI proved to be particularly effective in differentiating, being closer to 1 when pathology is not present.