Classification of mammography images based on cellular automata and Haralick parameters

摘要

This paper presents a bio-inspired algorithm that helps viewing image classes of mammography. This algorithm uses the concept of cellular automata (CA); the paradigm constructed from simple elements interrelated and form a regular structure with local neighbors' interactions. This connection generally builds a simple geometry; lattices, the vectors representing the database images located on 2 dimensions, and the local update rule of cells favoring the creation of similar states' groups in neighboring cells. This method helps to understand the underlying structures of all images. A possible solution for the classification of mammographic data by the cellular automata is promising. The results obtained from real images show the power of CA in classifying malignant and benign mammographic. Despite that, series of experiences evaluating this approach are described using the database provided by radiology department of Ibn Roshd hospital of Annaba in Algeria. From 180 images provided by the radiologist 79 are used (46 malignant and 33 benign). The results of applying this technique to mammography have been quite promising and are discussed in the following paper, the achieved sensitivity and specificity reach 89 and 84 %, respectively.