Texture Characterization for the Early Detection of Liver Diseases
with M.Popovic, S. Markovic and M. Krstic

The goal of this research was the detection of an optimal feature extraction technique for the description of different ultrasound textures. We have investigated the utility of wavelet decompositions as feature extraction methods in discrimination among diffuse liver diseases. We have applied a nonseparable quincunx transform and the traditional approach based on the separable wavelet transform, and compared them to the previously used techniques for texture characterization. Based on the experimental results we concluded that the quincunx transform (see the pictures bellow left) is provides lower sensitivity to rotation, and performs better in the characterization of noisy data. The results obtained in the study illustrate the possibilities of image processing in the ultrasound diagnostics of liver.

Texture Characterization of Infracted Myocardial Tissue
with A. Neskovic, M. Popovic, and A. Popovic

Some applications for tissue characterization in medicine and biology, such as analysis of myocardium or cancer detection operate with tissue samples taken from very small areas of interest. In order to perform tissue characterization in such applications, only few texture operators can be employed: these operators should be insensitive to noise and image distortion, and yet accurate in estimating the texture quality from the small number of available pixels. To describe the quality of the infracted myocardial tissue we proposed a new wavelet-based approach for analysis and classification of small texture samples. This method decomposes an image via the wavelet filterbank, and then computes image approximation on higher resolution (see the picture bellow right). Texture energy measures calculated at each output of the filterbank, as well as the energies of synthesized images are used as texture measures in the classification procedure. We also proposed an unsupervised classification technique based on the modified T-test. We tested the method with clinical data and achieved very promising results.

Nonseparable quincunx wavelet decomposition of a liver tissue sample (left) and wavelet based analysis of the infarcted myocardial tissue (right)

Related Publications
A. Mojsilovic, M. Popovic, "Characterization of visually similar diffuse diseases from B-scan liver images using the nonseparable wavelet transform", IEEE Trans. on Medical Imaging, vol. 17, no. 4, August 1998. pdf
A. Mojsilovic, M. Popovic, A. Neskovic, A. Popovic, "Wavelet image extension for analysis and classification of infarcted myocardial tissue", IEEE Trans. Biomedical Engineering, vol. 44, no. 9, September 1997. pdf
A. Mojsilovic, M. Popovic, R. Babic, M. Ostojic, "Automatic segmentation of intravascular ultrasound images: A texture based approach", Annals of Biomedical Engineering, vol. 29, November 1997.
A. Mojsilovic, A. Neskovic, M. Popovic, J. Marinkovic, M. Bojic, A. Popovic, "Texture analysis and classification with the nonseparable wavelet transform", Circulation, August 1998.
Back                                                                                                        Last modified on April 5, 2002.