Detection of Tumor in Digital Images of the Brain

This paper describes some achievements in the detection of tumor in medical images. A computer system has been designed and developed to recognize the typical features of the “glioblastoma multiforme” in the digital images of the brain. The basic concept is that local textures in the images can reveal the typical “regularities” of the biological structures. Thus, textural features have been extracted using a co-occurrence matrix approach. The analysis of the level of correlation has permitted to decrease the number of these features to the only significant components. An artificial neural network has been used for texture analysis and classification. The level of recognition, among three possible types of image areas: “tumor”,“non-tumor” and “background”, was characterized by a satisfying number of correct answers.