B-scan Image Identification Based On Fractal

Liver cancer is the most common malignancies.In China, hospitals mainly rely on doctors for medical image diagnosis. The diagnosis is very extremely intense and low efficient, and quantitative analysis is more difficult,and thereby it affects the accuracy of the clinical diagnosis.Computer-aided diagnosis for the early realization of the medical discovery of liver cancer is the urgent need to address the problem.Various internal organs graphics of the human body can be clear through B scan. As a result , B scan becomes the commonly used method in the hospital for discovering liver disease. But there are deficiencies identified by ultrasound examination. The resolution is not high enough, and some small lesions would not be able to be detected. Organ covered by some part of the high gas content is very difficult to demonstrate. So, we need a special image processing technology to the ultrasound image analysis and processing,to identify pathological features,and thereby making medical diagnosis.The fractal of images is self-similar,with no loss due to a variety of factors. Therefore,getting the fractal characteristics of the images can improve the recognition. And the ultrasound images can be viewed as a 3D coordinate plane whose pixel coordinates on the x-axis, y-axis coordinate and its z-axis values is gray values.So,B scan image and geological distribution have similar places,and we can use variogram to calculate the fractal value.First, the paper discrete the fractal Brown model,and on this basis,we find the relationship between the fractal characteristics and geostatistics variables (variogram) of the ultrasound image.Second,we design and implement the single-step methond to calculate the value of the variogram function,and get the fractal value of B scan images.Finally,we take the fractal values,the Fourier features, gray difference statistics and trip length statistics as classification attributes,and use vector support machine for classifying.Simulated annealing algorithm is used to search for the best support vector machine parameters, with the corresponding parameters of the support vector machine learning the samples. And then we get the support vector machine model , which is used in the image classification to determine whether there is liver cancer or not. Simulation results show that the fractal dimension in conjunction with other characteristics can be more effective in determine whethre there is liver disease or not.