Segmentação de lesões periapicais em imagens odontológicas radiográficas digitais utilizando modelos de contorno ativo com inicialização automática

Abstract

The image is used as one of the main diagnostic tools in several areas of health. In many cases, imaging-guided diagnoses are of paramount importance to reveal the cause of symptoms, to follow disease treatment as well as to define the right kind of treatment. For instance, the radiography, amongst the most popular medical imaging techniques, uses a beam of X-rays in order to create the image. When it comes to Dentistry, periapical radiographs play an important role in obtaining diagnoses of oral pathologies in patients. However, in the process of performing the diagnostic exams, the representation of the three-dimensional anatomical structures of the face is often created from two-dimensional images, damaging the interpretation by the dental surgeon. Moreover, the physical properties of lesions and tissues of the maxillofacial region may not be fully correlated with their anatomical limits, making segmentation of such lesions a very complex and specialized procedure, which often requires considerable manual interaction. The present study proposes the use of the Localizing Region-Based Active Contours, from an initial mask, to segment periapical lesions. Mathematical Morphology Techniques and Gabor filter are used to identify points for initialization as well as fine-tune the MCA’s initial mask adjustment. For a quantitative evaluation of the proposed approach, metrics such as Dice Coefficient, precision and accuracy are used to establish the proximity relation of the segmented region with the desired result.The average for Dice Coefficient was higher than 0:7 and standard deviation of less than 0:01. In general, the results were promising towards the accomplishment of the objectives and the progress of future studies in the area of image segmentation applied to health.