Texture-Based Methods and Dimensionality Reduction Techniques Involved in the Detection of the Inflammatory Bowel Diseases from Ultrasound Images

Abstract

The inflammatory bowel diseases (IBD) are severe, chronic and recurring disorders, requiring continuous patient monitoring. The most reliable methods for the diagnosis of the inflammatory bowel diseases are invasive (endoscopy, colonoscopy, histopathology) or irradiating (CT). We aim to develop computerized methods for the noninvasive assessment of the bowel inflammation level based on information obtained from ultrasound images. In this work, we study the role of the textural parameters in characterizing different types of inflammatory bowel diseases and the colorectal tumors. The dimensionality reduction techniques are taken into consideration in order to obtain the relevant textural features and to improve the result of the automatic diagnosis. The Principal Component Analysis (PCA) method and the Correlation based Feature Selection (CFS) method, as well as their combinations, are assessed for this purpose. The methods of Support Vector Machines (SVM) and Multilayer Perceptron (MLP), which gave very good results in our former experiments, are implemented for the automatic diagnosis. B-mode ultrasound images belonging to biopsied patients, are used. The patients were suffering from the following types of diseases: Crohn’s disease, ulcerative recto-colitis, colo-rectal cancer.