Fusão de modelos 3D com imagens térmicas para aplicações médicas

• Ano de defesa: 2015
• Autor(a) principal: Krefer, Andriy Guilherme
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Tecnológica Federal do Paraná; Curitiba; Programa de Pós-Graduação em Engenharia Elétrica e Informática Industrial
• Palavras-chave em Português: Imagem tridimensional em medicina; Processamento de imagens - Técnicas digitais; Imagens e fantasmas (Radiologia); Circuitos impressos; Métodos de simulação; Engenharia biomédica; Energia elétrica; Three-dimensional imaging in medicine; Image processing - Digital techniques; Phantoms (Radiology); Printed circuits; Simulation methods; Biomedical engineering; Electric power
• Link de acesso: http://repositorio.utfpr.edu.br/jspui/handle/1/1407

Thermography is an imaging method that allows temperature visualization of various regions of an object. In medicine, it finds applications related to oncology, burn trauma, vascular, respiratory and skin diseases, and as a general tissue vitality checking tool. 3D thermography adds tridimensional information to the conventional 2D thermography. It is made from a 3D mesh wrapped by thermal texture, enabling a more precise visualization of thermal patterns of anatomical structures. We propose a system capable of combining 2D thermal images with their corresponding 3D mesh, delivering a 3D thermogram for medical applications as a result. We used Particle Swarm Optimization (PSO) for mesh alignment and Structure from Motion (SfM) for 3D reconstruction in the present method. In contrast to most research found in the literature, the 3D mesh and the thermal images do not need to be acquired simultaneously, and a mechanical support for the thermal camera and the 3D scanner is not required. The 3D mesh may be acquired, for instance, from a 3D scanner or a magnetic resonance imaging machine. In order to evaluate the results, a phantom, that is, a static assessment object of known properties has been built. For this purpose, a novel technique using printed circuit boards has been developed. As a result, comparison between the output of the method and the phantom shows a maximum error of 3.73 mm and a mean error of 1.41 mm with 0.74 mm of standard deviation in phantom of 100 x 150 x 103.2 mm.