Alinhamento Automático de Imagens de Cintilografia Miocárdica Através da Combinação de Técnicas de Processamento de Imagens

摘要

The acute myocardial infarction (AMI) is considered the main cause of ischemic heart syndrome, one of the first causes of death in developed countries. One effective way to realize follow-up on patient that suffered AMI is the analysis of polar maps. This work presents an automated registration technique for myocardial scintigraphy images. The purpose of this algorithm is the positioning of the SPECT image according to a known pattern, allowing the automatic creation of bull's-eye maps for easier visualization. The developed method for the structure based registration utilizes different image processing techniques, like thresholding and feature extraction, to improve the result of the model based image registration. After the model based image registration, a technique for rotation adjustment in both long axes was applied, where the wall of the myocardium is detected based on its shape. Following, the technique for rotation adjustment in the short axis was applied, using information obtained through the SIFT technique. The proposed technique was appraised through comparison with the ideal alignment realized manually in 15 cases of patients with and without heart defect. The rate of success obtained was roughly 86% for the analyzed cases. The results obtained indicate that the proposed techniques were capable of realizing the registration automatically, even though using methods that have little computational costs.