Knowledge based segmentation, quantitative characterization and classification of basement membrane from oral histopathological images

Invasive growth of cancer cells is a complex process involving specific interactions between tumour cells and the orderly integrated complexes of the extracellular matrix. Basement membrane (BM) has been proposed as one constituent of extracellular matrix, which carries responsibility for regulating invasion and metastasis. In view of this, the research work presents a quantitative approach for analysis of thickness of the BM in oral submucous fibrosis showing dysplasia (OSFD), oral submucous fibrosis without dysplasia (OSFWD) and normal oral mucosa (NOM). In order to delineate the BM, we have developed a knowledge-based segmentation algorithm using anisotropic diffusion and fuzzy divergence based thresholding followed by colour based region growing. Further, we extracted the mean thickness of the BM. The significance of the extracted feature (thickness) is evaluated using statistical analysis it shows the feature is significant in discriminating the three groups; it is also observed that there is an increasing trend of BM thickness for OSFWD and OSFD compared to normal counterpart. Further, the significant features are fed to the support vector machine (SVM) classifier to discriminate (classify) normal, OSFD and OSFWD groups. The thickness feature provides a good sensitivity of 80.16%, specificity of 100% and positive predicative accuracy of 100%. Hence, it can be recommended as a quantitative biomarker in the context of histoptahological evaluation. This quantitative characterization of basement membrane will be of immense help for oral onco-pathologists, researchers and clinicians to assess the biological behaviour of OSFD and OSFWD, specially relating to their premalignant and malignant potentiality. As a future direction more extensive study involving more number of disease subjects is observed.