Ultrasonic tissue characterization using fractal feature

B-mode ultrasonic scanners are widely used in the hospitals and clinics for diagnosis. The image generated represents the ultrasonic appearance of the human tissues being examined. Tissue identification can then be carried out visually by the physicists. However, accurate interpretation the ultrasonic image can be difficult due to the possibility of human error and the defects inherent in the image. One solution is to carry out quantitative analysis on the ultrasonic image. Statistical texture analysis has been applied to many types of image to derive numerical parameters for quantitative characterization. For instances, first-order and second-order grey level statistical calculations have been commonly used. Recently, another approach for texture characterization begins to emerge, which is based on fractal geometry. The fractal model has the advantages that the parameters generated are stable over transformations of scale and linear transforms of intensity. One important parameter that can be derived is the fractal dimension. The fractal dimension can be determined using three separate methods. In the first method, the fractal dimension is estimated based on the average intensity difference of the pixel pairs within the region of interest. In the second method, the fractal dimension is determined from the Fourier power spectrum of the image data. The third method, also called reticular cell counting approach, estimates the fractal dimension by counting the number of subcubes enclosed by the intensity surface. In this investigation, quantitative characterization of ultrasonic image of carotid artery is carried out using fractal dimension and second-order statistical parameter. In each image, a number of regions are chosen for analysis. Each region corresponds to a particular type of tissue, for examples, blood, vessel wall, carotid plaque, etc. The effectiveness of each parameter is then assessed based on its ability to discriminate one type of tissue from others. In addition, the fractal dimension is compared with the second-order statistics in the segmentation of ultrasonic image. From the results obtained, the fractal dimensions estimated using the third method performs better than those obtained using the first and the second methods in tissue discrimination. The performance of fractal dimension, especially that obtained from the third method of reticular cell counting approach, is comparable to the second-order statistical parameter. For automatic image segmentation, the reticular cell counting approach is also as good as the second-order statistical calculation, whilst the former method has an advantage in faster speed.