Diagnosis Of Hepatic Fibrosis Based On Image Hardness And Edge Texture Features

Liver cirrhosis is a common clinical chronic liver disease, It is caused by a long-term variety of factors, and is a diffuse live damage. Liver fibrosis is the early clinical pathology symptom of liver cirrhosis, and has been found is reversible. Therefore, early detection and timely treatment of live fibrosis can significantly reduce the mortality rate of liver disease. At present, the diagnosis of liver cirrhosis and liver fibrosis mainly rely on biopsy methods. This invasive examination not only damages the physical and mental health of patients, but also the clinical experience of doctors will directly affect the results of diagnosis. Thus it is urgent to find a method of early non-invasive diagnosis of liver fibrosis.With the development of medical technology and the digital image processing technology, the computer-aided diagnosis (CAD) technology is introduced by people in the diagnosis of liver fibrosis, greatly improving the diagnostic efficiency and accuracy. According to the characteristics of the shape features is time-consuming and poor scalability and diagnostic effect of texture features is not good in the diagnosis of hepatic fibrosis, so in this paper, a research of computer-aided diagnosis of hepatic fibrosis on image hardness and edge texture features is proposed, the primary coverage of the study is arranged:(1) Analyzing the softness of the liver by image processing based on the grid shaped liver deformation image. In spatial domain Thin Plate spline (TPS) algorithm is used to calculate the minimum bending energy of deformation; in the frequency domain, Fast Fourier Transform (FFT) algorithm is used to calculate the spectral values of the characteristic region before and after the deformation. It was found that both the minimum bending energy and the power spectrum have a good linear relationship with the softness of the liver, and the joint distribution can distinguish the normal and abnormal liver.(2) Preprocessing the hepatic CT Images through Gray Level Co-occurrence Matrix and extract liver contour edge information:Mean Gray Value、Standard Deviation and 14 edge texture features, to do the classification diagnosis experiment of liver fibrosis by 16 edge texture features.(3) In hepatic fibrosis diagnosis classification experiment it takes support vector machine (SVM)algorithm based on radial basis function (RBF), to do classification and diagnosis experiment it takes the highest cost method of cyclic traversal method and Leave-one-out Cross Validation to loop through all feature combinations. The results of experiment show that optimal number of edge texture features is confirmed from 3 to 7; Analyzing of the weight ranking of the edge texture feature find that:the weight value of the maximum correlation coefficient, the contrast, the difference variance, the sum of square variance and the sum variance is far greater than other features in the classification experiment.(4) Comparing the experimental results of the edge texture feature and the edge texture feature、shape feature in the classification diagnostic of liver fibrosis; comparing the advantages and disadvantages of the feature extraction method; and analyze the characteristics of the feature weight distribution. Research on the liver fibrosis mixed diagnosis experiment based on image hardness and edge texture feature, and the overall average accuracy rate of the diagnosis was greatly improved by this mixed diagnostic experiment.