Possibilities of texture analysis of computed tomogram in diagnosis of chronic obstructive disease

Abstract

We analyzed the possibility to use different texture features for determination of the lung pathologic changes by examining two dimensional computed tomogram images. As informational features, we used the histogram and covariance characteristics as well as the Haralick features and the run length features. The effective features were selected basing on the discriminant analysis criterion. The obtained group of effective features provides only the part of 0.11 incorrect recognitions, which is better than the results obtained without using the procedure of the feature's selection.