TEXTURE DESCRIPTORS IN THE ANALYSIS OF CANCER IMAGES IN ANIMAL TISSUE

Authors

DOI:

https://doi.org/10.20873/saberesemcirculacao6

Abstract

In this work, we propose the separation of images of cancerous and healthy tissues obtained over time by combining machine learning methods with texture descriptors. Four sub-images were considered from each of the 128 images (covering healthy and cancerous areas), captured at a rate of 0.08 seconds, from a canine anaplastic mammary carcinoma, and another 128 images from a basosquamous carcinoma in the tail skin of a feline. For the analysis, Haralick texture descriptors were extracted from the images, and the most important descriptors for tissue classification were selected using the Random Forest method. Subsequently, the Independent Component Analysis (ICA) method was applied to these descriptors for each image under study. The weight matrix was then used for clustering the images. This methodology enabled the separation of sub-images of healthy and cancerous tissues, as well as distinguishing between images from cats and dogs.

Published

2025-12-19 — Updated on 2025-12-19

Versions

How to Cite

Safadi, T., Mendes Magalhâes Junior, A., & de Oliveira Pala, L. O. (2025). TEXTURE DESCRIPTORS IN THE ANALYSIS OF CANCER IMAGES IN ANIMAL TISSUE. DESAFIOS - Revista Interdisciplinar Da Universidade Federal Do Tocantins, 12(7), 266–279. https://doi.org/10.20873/saberesemcirculacao6