Histopathological Image Based Oral Pre Cancer Grading Using Machine Learning

Early detection of oral cancer can be achieved through pre-cancer grading. It is an important task in prescribing necessary treatment and medication for the patients. Due to its high mortality and morbidity rates, oral cancer tends to be discovered at a later stage and is fatal to the patients. The pathologists manually stain and analyse the histopathological samples of the patients and it is a laborious and time-consuming process. Artificial intelligence approaches have a considerable impact in improving diagnostic accuracy in all fields of medicine. The oral pre-cancer features extracted for grading are cell nuclei size, nuclei intensity i.e. hyperchromasia and cytoplasmic ratio from real-time dataset of histopathological images. The results are compared using several machine learning algorithms, out of which, random forest has an accuracy of 84.2%. This research aims to develop a diagnostic tool to aid medical practitioners for automatic and fast pre-cancer grading and establish the importance of early diagnosis of the disease.