Quantitative Analysis of Sub-Epithelial Connective Tissue Cell Population of Oral Submucous Fibrosis Using Support Vector Machine

This work presents a quantitative microscopic approach for discriminating inflammatory and fibroblast cells of oral submucous fibrosis (OSF) from normal oral mucosa (NOM) in respect to shape features of the sub-epithelial connective tissue (SECT) cells. However, malignancy develops only in the epithelium; significant pathological changes are evident in the SECT concurrently. The changes in SECT cell population will spell the intricate biological behaviour pertaining to normal cellular functions as well as in premalignant and malignant status. In view of this, the present work characterizes the SECT cells (inflammatory and fibroblast) using their shape parameters. In this study segmentation and classification of sub-epithelial connective tissue (SECT) cells except endothelial cells in oral mucosa of normal and OSF conditions has been reported. Segmentation has been carried out by colour deconvolution and subsequently the cell population has been classified using Support Vector Machine (SVM) based classifier. Moreover, the shape features used in this study are statistically significant using Mann Whitney U test, which enhance the statistical learning potential and classification accuracy of the classifier. Automated classification of SECT cells characterizes this precancerous condition very precisely in a quantitative manner and unveils the opportunity to understand OSF related changes in cell population having definite geometric properties. The paper presents an automated classification method for understanding the deviation of SECT cell population from normal to precancerous stages. The SVM classifier is trained and tested with 15 features for the classification between normal and OSF samples. Experimental results are obtained and compared. It is observed that linear kernel based SVM identifies the unknown SECT cells with an average accuracy of 96.55% for normal, 94.65% for OSFWD and 92.33% for OSFD groups. This quantitative characterization of SECT cell population will be of immense help for oral onco-pathologists, researchers and clinicians to assess the biological behaviour of OSF, specially relating to their premalignant and malignant potentiality.