Texture Analysis of Medical Images Using the Wavelet Transform

Texture analysis of images can contribute to a better interpretation of medical images. This type of analysis provides not only qualitative but also quantitative information about tissue affection degree. In this work an algorithm is developed which uses the wavelet transform to carry out the supervised segmentation of echographic images corresponding to injured Achilles tendon of athletes. To construct the pattern, the image corresponding to healthy tendon tissue of the athlete, is taken as a reference based upon the duplicity of this structure. Texture features are calculated on the expansion wavelet coefficients of the images. The Mahalanobis distance between texture samples of the injured tissue and pattern texture is computed and used as the discriminating function. It is concluded that this distance, after appropriate medical calibrations, can offer quantitative information about the injury degree in every point along the damaged tissue. Further, its behavior along the segmented image can serve as a measure of the degree of change in tissue properties. The parameter, similarity degree, is defined and obtained by taking into account the correlation between distance histograms for the healthy tissue and the damaged one. It is also shown that this parameter, when properly calibrated, can offer a quantitative global evaluation of the state of the injured tissue.