Colorectal cancer recognition from ultrasound images, using complex textural microstructure cooccurrence matrices, based on Laws' features

The colorectal cancer is a frequent lethal disease nowadays. The most reliable method for diagnosis, the biopsy, is invasive and dangerous. We developed computerized, texture-based methods, for non-invasive cancer diagnosis, using the information obtained from ultrasound images. In this work, we defined the co-occurrence matrix of complex textural micro-structures, determined by using the Laws' convolution filters, in conjunction with clustering methods and we experimented it in order to perform automatic diagnosis of the colorectal tumors. These tumors were compared with the Inflammatory Bowel Diseases (IBD), as they have the same appearance with these affections in ultrasound images. We also determined the relevant textural features that characterize the colorectal tumors, by using specific methods. For the automatic recognition, we used powerful classifiers, such as the Multilayer Perceptron (MLP), the Support-Vector Machines (SVM), decision-trees based classifiers such as Random Forest (RF) and C4.5, respectively AdaBoost in combination with the C4.5 algorithm.