Reconhecimento de nódulos pulmonares utilizando a transformada de Hough e textura

• Autor(a) principal: Severiano, Fabio da Silveira
• Data de Publicação: 2015
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da UERJ
• Texto Completo: http://www.bdtd.uerj.br/handle/1/13693
• topic: Detection; Lung Nodules; CAD; Hough Transform; Texture Analysis; Pulmões - Doenças; Pulmões Doenças Diagnóstico; Pulmões Câncer; Processamento de imagens auxiliada por computador; Detecção; Nódulos Pulmonares; CAD; Transformada de Hough; Análise de Textura; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO

This project demonstrate the eficiency of the automatic detection of solitary pulmonary nodules, using the geometry of the nodules, which is elliptical or oval, and the texture analysis. The Hough Transform, is able to locate circular and elliptical structures in an image and texture descriptors which indicate the characteristics of a given structure in an image. This process have five steps: the location of the lungs, the location of candidate structures, characterization of structures, database creation and recognition of the nodule. The location of the lungs fase consists in the reduction in the image noise and its equalization. Then the image is binarized, eroded and have your structures filled with a morphological fill. Because this all structures except the lungs are removed. The detection of candidates for nodules starts with the segmentation of existing objects in the lung. These structures are turned into edges by the Canny filter and has its geometry analyzed by the Hough transform. In the Caracterization stage, the relevant characteristics of the texture of the candidates are analyzed. First is calculated the Co-occurrence Matrix in the region, which stores the gray level transitions, from neighboring pixels, at angles of 0°, 45°, 90° and 135°. Then, the system find the probability of each transition occur and calculates the following texture features: Maximum Probability, Second Moment Angular, also known as Energy; Entropy, contrast, correlation, homogeneity and Variance The obtained values are compared using the nearest neighbor technique with a previously registered database with classified elements as nodules and other structures, recognizing the nature of the structure. This technic have 70% of efficience in the classification and in 99% of the cases, the Circular Hough Transform had selected the nodule.