Edge Block-Based and Support Vector Machine for Oral Cancer OCT Image Segmentation

• Main Authors: Tsung Chin Chen, 陳宗琴
• Other Authors: J. D. Lee
• Format: Others
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/06879734122627020165

碩士 === 長庚大學 === 電機工程學系 === 100 === Optical Coherence Tomography (OCT) has becoming a new tool for diagnosing oral cancer in recent years. However, due to the scattering characteristics, speckle noise usually exists in the obtained OCT image by receiving random function energy. In order to solve the problem, this study uses median filter to eliminate the speckle noise, and then the edge blocks are employed as the training set for a SVM. Moreover, to validate the performance of this proposed method, we utilize different type of OCT images, such as the image of oral cancer, skin surface, etc, as the material in the experiment. The classifier trained with edge blocks from these OCT images are used for image segmentation. More specially, the segmentation of oral cancer image is a pixel-based approach by predicting each pixel using a trained classifier. After segmenting the edge of various tissues, the distance between two tissues can be obtained and it is defined as the depth of a tissue layer. The depth change of the tissue layer before and after curing is useful for medical doctor to evaluate the patient condition. For skin image, segmentation technique is used to label the edge of skin surface and then calculate the area with thermal damage due to laser. The experimental results show the proposed method can effectively segment various OCT images obtained from clinical applications.