Software system for the automatic and computer assisted diagnosis of some severe abdominal affections, based on ultrasound images

In this article we describe specific methods for the characterization and computer assisted diagnosis of some severe abdominal diseases, based on ultrasound images and the corresponding software system. The objective of our research is to build an instrument for the non-invasive diagnosis, based on virtual biopsy, of some severe abdominal diseases such as the cirrhosis, the hepatocellular carcinoma (HCC) and the colorectal tumors. The gold standard for the diagnosis of these affections is the classical biopsy, but this is invasive, dangerous. Ultrasonography is a medical-imaging based investigation method that is noninvasive, inexpensive, safe and thus repeatable. Texture is an important property of the human body tissues, able to reveal subtle characteristics upon the pathology. We defined the imagistic textural model of these affections, using the information obtained from ultrasound images. Based on the imagistic textural model, we assessed the possibility of performing automatic recognition of the considered affections both in supervised and unsupervised manner, the purpose of the latter approach being that of discovering the severity grades of the diseases. The resulted accuracy was above 90% in most of the cases.