Abstract
As the quest for curing cancer continues with ever growing determination and vigor, new medical imaging techniques have come up to help in early cancer detection. The early detection aids in the possibility of either complete cure or at least lengthening the life of the affected individual. Elastography is one such recently developed image acquisition technique. It maps the ultrasonically estimated computed strains of the object under observation to an image, corresponding to an applied stress. The images are known as elastograms. Since tissue stiffness is one measure of tissue abnormality, imaging the elastic properties of the tissue would bring important and useful information about the tissue. The work, still in research phase, is being evaluated and developed to make it a commercial product. 'This research applies image texture analysis to computer generated elastograms to obtain effective texture features. Four image analysis techniques, no-occurrence statistics, wavelet decomposition, frontal analysis and granulometry are used to extract a number of features from each image. The inclusions in the elastograms simulate real tissue abnormality. The aim is to find effective features that can track the underlying parameters of elastogram generation like hardness and density with the feature distributions being as apart as possible. Fisher discriminant is used as a statistical separability measure to find effective features for various cases of density and hardness differences. Among the features studied, mean and wavelet features are found to be effective in differentiating inclusions having different stiffness or density.
Hussain, Fasahat (1999). Image texture analysis of elastograms. Master's thesis, Texas A&M University. Available electronically from
https : / /hdl .handle .net /1969 .1 /ETD -TAMU -1999 -THESIS -H87.