Detrimental To Liver Ultrasonic Images Based On Texture Analysis To Identify

In this thesis, by applying computer technology ,biomedicine technology and pattern recognition, a methodology about classification of ultrasonic liver images is proposed by the texture features. Hence I introduced the Bayesian framework to the image recognition about live lesions. The method combined the co-occurrence matrices with Multiresolution fractal feature. First, according to the usage of image's texture in pattern recognition, it is include the apatial gray-level dependence,　the Fourier power spectrum, the gray-level difference statistics and Laws'texture energy measures that some feature have used in distinguishing threes kinds of ultransonic live images:　normal liver, cirrhosis, liver cancer. And the Bayesian classifier is employed to distinguish. Second, the concepts of multiple resolution imagery and fractional Brownian motion is proposed to detect diffuse liver disease fastly. A new feature set ——the multiresolution fractal features are used to improve the speed and accuracy of classification in this thesis. Third, I used 60 samples (30sample each) to distinguish these images according to a block of 32*32 pixels. I adopt statistical pattern recogntion according to the trait of ultransonic image and liver's texture. And I optimized Baysian classifier using parametic learning. Last I had united all data in order to finishing finial class. And I had a great many analysis and explanation for each feature. A real time implentation of our algorithm is performed on a personal computer and is to evaluate the performance of these features in addition. The experimental results show that this method is firstly applied to the classification about three kinds of ultransonic live images.　The correct rate is 87.48% which is satisfying. Accoring as my country situation at present, The study have greatly factual significance.The method provides strong theoretical basis for the hospital without consultant and the telemedicine information system, also provides the theoretical support to the computer aided diagnostic system about live lesions.