Abstract

Oral cancer is one of the most prominent diseases across the world with an estimated 657,000 new cases each year, and more than 330,000 deaths. Due to the vast consumption of tobacco and alcohol in India, over one lakh cases of oral cancer have been reported in 2018. The treatment of oral cancer is generally effective if it is diagnosed at an early stage. The application of image processing for diagnosis of on-site patients is an effective way to identify precursor oral lesions without performing any unnecessary biopsies. Due to the observer bias, computerized analysis of biomedical images has become an important research area. In this paper, we present a method to automatically extract oral lesion area and generate a corresponding binary mask. The dataset recorded contains digital pictures of marked and unmarked oral lesions. The experimental results show that our proposed method performs the task of lesion segmentation with high accuracy. The obtained mask is then being used for classification process using machine learning techniques.