CNN based automatic tongue cancer detection using hybrid k-mean and watershed transformation

  • Ms. Pallavi Pahadiya, Dr. Ritu Vijay, Ms. Shivani Saxena, Mr. Kumod Kumar Gupta, Dr. Tushar Shahapurkar
Keywords: Tongue Diagnosis system (TDS), K-mean clustering, Firefly Algorithm, Particle swarm optimisation, Watershed transform, CNN

Abstract

In India oral cancer is becoming community health issue. Due to the changing life style and increase in fear due to contact diseases like covid-19, there is increase in demand for non-invasive methods for disease identification. Research is going on to identify oral cancer using portable devices which can reach remote places and can decrease the rate of mortality by identifying cancer at early stage. Digital camera and mobile based camera are available everywhere and can cover interior of villages as they are portable and less costly. Cost of dedicated costly hardware used for Tongue image identification can be tried to overcome this issue. This paper presents digital tongue image based analysis using multi stage segmentation. For segmentation hybrid Particle swarm fireflies based k-mean clustering is proposed for initial segmentation followed by watershed transformation to get minor abnormal areas.CNN with augmentation is used as a classifier. Online available images and 150 normal, abnormal and cancerous images are used which gives 90% validation accuracy. It is able to identify the non cancerous and cancerous segments of input image.

Published
2021-07-06
How to Cite
Mr. Kumod Kumar Gupta, Dr. Tushar Shahapurkar, M. P. P. D. R. V. M. S. S. (2021). CNN based automatic tongue cancer detection using hybrid k-mean and watershed transformation. Design Engineering, 1864- 1879. Retrieved from http://thedesignengineering.com/index.php/DE/article/view/2507
Section
Articles