Feature Extraction And Analysis For CT Image In High Incidence Of Hepatic Hydatid Disease In Xinjiang

Objective: To get the feature of strong classification ability, feature extraction andfeature analysis had been applied in CT image in high incidence of hepatic hydatid diseasein Xinjiang. Then further explore these features application in image classification ofhepatic hydatid diseases, to lay the foundation for content-based medical image retrivalsystem in high incidence of hepatic hydatid disease in Xinjiang. Methods: Through theMatlab image processing software, firstly, processing of CT image, to improve the qualityof the image, saves the valid information and deletes unnecessary information. And thenextract the feature of the processed image based on the gray-scale histogram, GLCM andkolmogorov complexity. Through the SPSS statistical analysis software, using maximumclassification distance method and significance analysis for feature analysis and accordingto the results of the feature analysis composed the comprehensive feature of the image.And, furthermore, using these features for classifies the image in high incidence of hepatichydatid disease in Xinjiang. Results: Using the maximum classification distance methodto analyze features of gray-scale histogram, GLCM and kolmogorov complexity. Thenapplied feature analysis results of gray-scale histogram, GLCM and the composedcomprehensive feature to classify the image. The result showing that the correctclassification rate between the image of healthy liver and unilocular hepatic hydatid cystare81%and71%,85%and66%,91%and87%, respectively. The correct classificationrate between the image of healthy liver and multiple daughter hepatic hydatid cysts are89%and82%,81%and72%,90%and93%. The correct classification rate between theimage of unilocular hepatic hydatid cyst and multiple daughter hepatic hydatid cysts are75%and74%,75%and76%,85%and80%. Using the significance analysis method toanalyze features of gray-scale histogram, GLCM and kolmogorov complexity. Thenapplied feature analysis results of gray-scale histogram, GLCM and the composed comprehensive feature to classify the image. The result showing that the correctclassification rate between the image of healthy liver, unilocular hepatic hydatid cyst andmultiple daughter hepatic hydatid cysts are84%,58%and77%;82%,77%and87%;96%,86%and86%, respectively. Conclusion: Image feature extraction method has beensuccessfully introduced into analyzes of CT image in high incidence of hepatic hydatiddisease in Xinjiang, for extract the feature and analyzes the feature. Through the analysisresults generated the comprehensive feature of image, which has higher correctclassification rate than the single-feature and meet the classification requirements to someextent. The results can be further been applied to content-based medical image retrievalsystem of high incidence of hepatic hydatid disease in Xinjiang and has a certain positivevalue.