Integrated phenotyping employing image texture features

Abstract

Image texture feature values are computed for a set of image texture features from an image of an anatomical feature of interest in a subject, and the subject is classified respective to a molecular feature of interest based on the computed image texture feature values. The image texture feature values may be computed from one or more gray level co-occurrence matrices (GLCMs), and the image texture features may include Haralick and/or Tamura image texture features. To train the classifier, reference image texture feature values are computed for at least the set of image texture features from images of the anatomical feature of interest in reference subjects. The reference image texture feature values are divided into different population groups representing different values of the molecular feature of interest, and the classifier is trained to distinguish between the different population groups based on the reference image texture feature values.

Dependencies are inferred from the claim text.