Optical quantitative pathology of cervical intraepithelial neoplasia in human tissues using spatial frequency...

An optical quantitative histological method in human tissues using spatial frequencies is demonstrated. Optical spatial frequency spectra from different stages of human Cervical Intraepithelial Neoplasia (CIN) tissue are evaluated as a potential quantitative pathological tool. The degree of randomness of tissue structures from normal to different stages of CIN tissue can be recognized by spatial frequency analysis. The standard deviation, σ of human normal and CIN tissue, is obtained by assuming the spatial frequency spectra as a Gaussian distribution. A support vector machine classifier (SVM) is trained in the subspace of σ. Twenty-eight normal and CIN samples of varying grades are examined and compared with current diagnostic outcomes. Our results suggest that an excellent accuracy for diagnostic purposes can be achieved. This approach offers a simple, efficient and objective way to supplement histopathology in recognizing alterations from normal to different stages of cervical pre-cancer, which are reflected by spatial information contained within the aperiodic and random structures of the different types of tissue.