Neighborhood Feature Extraction and Haralick Attributes for Medical Image Analysis: Application to Breast Cancer Mammography Image

This paper describes a new preprocessing method using Markovian modification to distinguish preprocessed mammograms (healthy and pathological), and these techniques are very necessary to find the direction of mammograms so that we can eliminate noise, improve image quality and obtaining more appropriate images than the original images as well as the disappearance of misplaced discrete points. Markovian image segmentation was performed to extract the region of the blocks if the blocks ended up being divided into groups of pixels that are homogeneous with the image with respect to certain criteria, based on the neighborhood system, using a relaxation approach that It is maintained by iterative conditional patterns (ICM), relying on an energy function adapted to the irregular neighborhood models of the image, and with the energy criterion of the image groups, the neighborhood system, and some cluster groups. These preprocessed mammograms are analyzed using co-occurrence matrices from which Haralick traits are extracted. The peculiarity of this approach is that it selects features from among the most selective mammograms based on the type of contact chosen for preprocessing. The most discriminating features are selected according to a supervised scheme that makes it possible to represent mammograms in a relatively small space where they can be discriminated with high reliability. To test this method and verify its validity, we use mammograms of normal and sick cases from the Reference Center for Reproductive Health in Kenitra, Morocco (CRSRKM).