Research Of Computer Aided Diagnosis Based On Liver CT Images

At present,due to the large number of liver diseases in our country,the disease rate is increasing,people's life and health is seriously threatened.But because of the particularity of human liver,liver pathological features of early is not very obvious,so the patients cannot find symptoms in early HCC,once the symptoms of treatment,is already late in the treatment after the diagnosis is very difficult,but the effect is poor,the survival time is short.To this,liver disease computer aided diagnosis(CAD)came into being,in the actual clinical disease census provides quantitative diagnosis,provide a reference for the diagnosis according to the doctor,has a positive impact is essential for early diagnosis of liver diseases.In this paper,after the deep research of the computer aided diagnosis technology,the abdominal CT image is used as the data source,and the main work is as follows:The first study is to improve the liver segmentation algorithm,proposed the use of SFCM two confidence connected layer segmentation method on liver segmentation.Based on preprocessed image to improve the quality of CT image,the first use of SFCM algorithm for CT image initial liver segmentation,image segmentation and the use of success the automatic generation of second confidence seed points connection algorithm,and the restriction of Canny again accurate segmentation,finally filling method make up the image.Through qualitative and quantitative analysis of the experiment,method validation improved liver segmentation sensitivity and efficiency,reduce the error rate.The average sensitivity is as high as 87.65%,which is improved by about 5%compared with the traditional segmentation algorithm.Finally,the three-dimensional model is reconstructed using the Marching Cubes algorithm,and the segmentation effect of the improved algorithm and the traditional algorithm is compared.The results of the three-dimensional reconstruction are analyzed.Second study of liver features,is proposed to extract the mixed feature based on wavelet transform,in order to reduce the computation,improve the extraction method of the feature of gray level co-occurrence matrix,and optimized by genetic algorithm for feature classification,streamline statistical characteristics were obtained by way of better.Through the classification accuracy and ROC curve analysis results can be obtained:the accuracy of abnormal liver classification is as high as 96.7%,the liver cyst resolution is as high as 92.6%,liver hemangioma and liver cancer classification accuracy rate of up to 80.3%.It has higher classification accuracy than the first order statistical feature,gray level co-occurrence matrix and all the features that are not selected.