Identification and Classification of Oral Cancer Using Machine Learning Techniques

In recent time, oral cancer is one of the major health issues in the world [1] and it is most pervasive in the middle- and low-income country. Early detection process of this disease is much essential for beginning the treatment procedure for oral cancers. Convolutional Neural Network” (CNN) model is the primary process that assist to detect the cancer through datasets such as Common object in context, visual objects classed. Accordingly, CNN helps to build a framework for detecting the object. it also uses for the classification of the images. This disease detection process can be prevented by early treatment and can save many lives and as the symptoms are mild it is tough to differentiate from other minor diseases. A better screening process and advanced detection can cure cancer and decrease the rate of cancer. This particular study shed lights on the entire classification and identification process of oral cancer. There are two kinds of method such as automated classification and automated detection process of oral lesion that are based on the deep learning-based detection frame work and deep learning-based classification frame work. This study aims for the recognition of different types of oral cancers, especially buccal mucosa cancer (cancer in the inner cheek). Machine learning algorithms and deep learning [10] processes can help in the accurate detection of oral cancers as well as buccal mucosa cancer. There are several elements need to be properly implemented for the success of the machine learning process. In this study CNN model and VGG-16 used for the identification and detection process. The data used for the process often accommodates many unfavorable factors, such as noise, blurry pictures or angle of click, that are supposed to be drawn out before the processing. Histopathological and normal camera images are used in the screening process, and data preprocessing before the screening process should be done. On the other hand, feature extraction, data cleaning, data normalization, data transformation and selection. This processed data is used as the final data set and enhances the screening process. This study aims for accurate data, so the “Convolutional Neural Network” (CNN) model, the method of machine learning [1] and the deep learning approach is used in this study. Biopsy images, histopathological images and camera images for cancerous and non-cancerous lesions are used separately in the model for better results. This model provided the most accurate results with the help of a transfer learning approach with hyperparameter tuning. A sum of 164 histopathological and camera images were used for the model. From this study it has been seen that the model provides accurate and best outcomes from datasets. The feature extraction and hyperparameter tuning increased the rate of accuracy level of detection process.