Discrete principal component analysis (DPCA) for pattern recognition and diagnostics of tissue pathologies such as cancer

摘要

A method for recognizing figures in an image frame provided by an optical device is disclosed. The method includes selecting a set of calibration images having characteristic figures; selecting a number of horizontal and vertical channels for dividing the image, obtaining a digital signature for the image corresponding to each channel to create a calibration image matrix, obtaining a first set of characteristics and a second set of characteristics from said calibration image matrix and selecting two coordinates from at least one component in each of said first and second sets of characteristics from said calibration image matrix; forming a pattern vector from said selected coordinates from at least one component in each of said first and second sets of characteristics from said calibration image matrix, forming a calibration pattern matrix using pattern vectors from a plurality of said calibration images, forming a sample pattern vector from said selected coordinates from at least one component in each of said first and second sets of characteristics from a sample image, obtaining a two-dimensional projection of the vector resulting from transforming a sample pattern vector by the calibration pattern matrix, and determining whether the sample figure corresponds to a calibration pattern by the location in a two-dimensional plot of said projection. A method for recognizing tissue conditions in a sample tissue image frame provided by an optical device using the pattern recognition as described above is also provided.