Image Segmentation Systems
In this context we have worked in two different projects. The first one deals with semi automatic tool for teeth segmentation in CT imagens. The main goal in this case is to allows the a three-dimensional reconstruction of tooth structures. Using the developed system it is possible to do automatic and manual segmentation from DICOM images and generates as output a 3D file. This file can be loaded in CAD (Computer Aided Design) systems that are able to create the three-dimensional models from plane sections. Figure 19 shows the some imagens from the system.
The second project in this subject deals with lung segmentation and automatic detection of lung tumors. Lung diseases victimize a huge amount of people by year. Manual diagnosis requires a vast number of tomographical images analysis, by radiologists. An alternative to this is the computer’s aid to this large scale of analysis, possibly related to subjectivity due to emotional factors. The computer provides a “second-opinion” to the radiologist, notifying him in case of something suspicious in the analysis. The present work proposes a computational system that uses co-occurrence matrixes descriptors, for lung diseases aided detection. Using these informations, the system is instructed with the values of the intervals of these descriptors, for each pattern category provided by the doctor. These values consist in the basic rule to define if a lung region is highlighted or not, in a thorax image. In the final tests execution, the capacity of the system, as a proposal of aided detection of lung diseases, was evaluated, and the results were satisfactory. Figure 20 shows the some imagens from the system.