Quantitative Analysis For Fatty Liver Medical Ultrasonic Images

Medical ultrasonic imaging, computerized tomography, magnetic resonance imaging and nuclear imaging are the four important medical imaging techniques in modern society. Among them, medical ultrasonic imaging is widely used in medical diagnosis and treatment due to its unique characteristic of real-time, noninvasiveness, cheapness, easy repetition, and high sensitivity. It has high potential in quantitative analysis, real-time monitoring and treatment scheduling, etc. Based on its virtues, it is often used to diagnose the disease of fatty liver.Nowadays, the diagnosis method for fatty liver, based on medical ultrasonic imaging, are that the doctors consults the basic metrical standard of ultrasonic machine and by right of their experiment to evaluate and diagnose the fatty liver disease. Because it lacks the quantitative evaluation mechanism, and has not enough warranty, misdiagnosis often occurs, and that will take the spirit burden to the sick. This thesis uses the texture analysis technique, belonging to the modern image processing domain, to analyze the medical ultrasonic images, and get the feature parameters. Then using these parameters to do the statistical analysis, and getting the quantitative data, comparing with the animal pathology results, at the end, this thesis gives the quantitative evaluation system for different degree fatty liver disease, That will change the actuality about the ultrasonic diagnosis.Recently, there are many image processing algorithms can deal with the classification problem of fatty liver medical ultrasonic images, but this problem is about the two classes analysis, and there is no method can give the quantitative data to describe fatty liver medical ultrasonic images. But this paper not only adopts the texture analysis algorithm to distinguish the three classes, but also uses the pattern reorganization and statistics to give the quantitative evaluation system. This quantitative evaluation system can be uses to change the traditional analysis method that can give the qualitative analysis, but not quantitative analysis.In order to provide abundant image resource and scientific evaluation warranty for results, the animal model experiment is made in Dalian Medical University's Experiment Animal Center. After getting the fatty liver animal model, the examination with the B-mode ultrsonography is done, and then getting these animal models' livers to do the pathological analysis. These pathological analysis results will be used to provide the impersonal warranty for this thesis's quantitative evaluation system. This thesis first gives an overview of the medical image texture analysis method. Then several commonly texture analysis algorithms are analyzed with reference to the experimental results on the ultrasonic images, and gives the qualitative result for classifying the different degree fatty liver disease. Because these methods cannot give the quantitative evaluation, so as to complete this thesis's target, this thesis carries on the work in following four aspects. Firstly, in order to quantitatively analysis the liver ultrasonic images well and get the quantitative data correctly, this thesis proposes a pretreatment method for the liver ultrasonic images, namely Peak-and-Valley filter. This method not only can rid of noise, but also can keep the edge information of images. Secondly, this thesis proposes a quantitative evaluation algorithm, a quantitative analysis system of fatty liver medical ultrasonic based on multiresolution fractal Brownian motion model and genetic algorithm, to give a quantitative evaluation system to diagnose the fatty liver disease. Thirdly, in order to give the visual quantitative evaluation system, this thesis proposes a method, fatty liver medical ultrasonic images quantitative analysis based on Gabor filter and bispectrum. From the bispectrum diffusion degree, this thesis gives the visual and intuitionistic results. Fourthly, using three texture analysis algorithms to analyze the human fatty liver medical ultrasonic images, and giving the conclusions about those analysis results.