Grading of Dysplastic Stages of Cervical Tissue through PCA Analysis of Mueller Decomposed Images

Abstract

Here we report the results derived from principal
component analysis (PCA) of Mueller decomposed images of cervical
tissues. PCA multivariate analysis, depolarization power data is
reduced to a few important components well suited for diagnostic
purpose. The subtle changes in the polarization property of the tissue
medium, with the progression of dysplasia, are found to be well
captured by the second and third principal components of the
covariance matrix, which clearly differentiate the normal and diseased
tissues. Interestingly, the principal components are found to be
sensitive enough to discriminate between two different stages of
dysplastic conditions. Hence, PCA technique allows automated
diagnosis that enhances the potential of real-time clinical detection of