IDENTIFICATION AND CLASSIFICATION OF ORAL CANCER LESIONS IN COLOR IMAGES

Abstract

Oral Cancer is the most common form of cancer in India. Poor adult villagers from remote areas in their 60s and 70s are the usual victims of oral cancer. The peculiar nature of oral cancer is that it is curable if treated properly at right time. In India this is particularly deadly due to the extensive use of tobacco coupled with lack of proper diagnostic facilities. A biopsy done at right time could save many lives. But it is the ignorance regarding the right time that causes the consequent la sufferings and unavoidable death sentence. This happens since the symptoms appear to be very much similar to a normal mouth ulcer and resulting negligence and complacency. This underscores the need for easier methods to identify this villain in disguise at an early stage. Early diagnosis via mass screening initiatives in rural areas with the help of image classification systems can help to reduce the mortality, medical costs, pain and sufferings. True color images are the most easily available input to such a system. The separation of mouth cavity images into cancerous and non-cancerous is a three stage process involving segmentation (Identification of ROl), feature extraction and classification. The objectives of this work include: Study of different methods for segmentation, feature extraction and classification. Selection of methods suitable for the present purpose. Application of the selected methods on the database. The test database is formed by sample images acquired from the Medical College under HIHT, Dehradun. Proposing the best system suitable for classification of images into cancerous and non-cancerous.