Segmentation and analysis of surface characteristics of oral tissues obtained by scanning electron microscopy to differentiate normal and oral precancerous condition.

Abnormal epithelial stratification is a sign of oral dysplasia and hence evaluation of surface characteristics of oral epithelial region can help in detection of cancerous progression. Surface characteristics can be better visualised by Scanning Electron Microscopy (SEM) in comparison to light microscopy. In our study we have developed automated image processing algorithms i.e. Gaussian with median filtering and Gradient filtering, using MATLAB 2016b, to segment the surface characteristics i.e. the ridges and pits in the SEM images of oral tissue of normal (13 samples) and Oral Submucous Fibrosis (OSF) (36 samples) subjects. After segmentation, quantitative measurement of the parameters like area, thickness and textural features like entropy, contrast and range filter of ridges as well as area of pit and the ratio of area of ridge vs. area of pit was done. Statistical significant differences were obtained in between normal and OSF study groups for thickness (p=0.0107), entropy (p<0.00001) and contrast of ridge (p<0.00001) for Gaussian with median filtering and for all the parameters except thickness of the ridge(p=1.386), for Gradient filtering. Thus, computer aided image processing by Gradient filter followed by quantitative measurement of the surface characteristics provided precise differentiation between normal and precancerous oral condition.