Criação de uma metodologia computacional para a segmentação bimodal de patologias timpânicas

by Junior, Vilson Heck

Abstract (Summary)
This work presents the development of a computational methodology for the bimodal segmentation of tympanic pathologies in digital video-otoscopic images. Based on the needs exposed by the Brazilian Chronic Otitis Media Center, where several video-otoscopic images are produced daily, the need to computationally quantify such pathologies arose. Video-otoscopic images possess a low contrast and a limited set of colors. These factors make the development of a semiautomated methodology more feasible than the development of a completely automated methodology. After the video-otoscopic image has been segmented, one can measure the pathologies that were segmented. This measuring allows the follow-up of the pathology and the treatment thereof. A method for the definition of the tympanic area was devised so that the ratio between the pathology area and the tympanic area can be computed. The results obtained by using the methodology underwent two similarity-based validation processes, Rand and BGM. They were also subjected to calculations aiming at determining the percentage of the pathology areas that was misclassified and thus yielded false-positives or false-negatives.
This document abstract is also available in Portuguese.
Bibliographical Information:

Advisor:Silvia Modesto Nassar; Eros Comunello; Sady Selaimen da Costa; Olga Regina Pereira Bellon

School:Universidade Federal de Santa Catarina

School Location:Brazil

Source Type:Master's Thesis

Keywords:ciência da computação processamento de imagens patologia otite média crônica ciencia computacao


Date of Publication:11/28/2008

© 2009 All Rights Reserved.