Research On Medical Image Segmentation And Lesion Extraction

Image segmentation is an important research topic in digital image processing,pattern recognition and other fields.Especially,it plays an increasingly important role in medical image.As the traditional medical image segmentation is basically based on artificial segmentation,segmentation results are often unsatisfactory and time consuming.Therefore,it has been the focus of medical image processing to achieve the automatic image segmentation and recognize basic lesion image.In this paper,we consider the combination of the active contour method and the maximum interclass variance Otsu method for the segmentation without considering the nonuniformity of the gray information distribution and the fact that some details can not be well segmented.Considering the texture features and shape features of the lesions,we classify lesion images by using support vector machine(SVM).The main contents are as follows:(1)Study on medical image segmentation.Due to the complexity of various organs and tissues in medical segmentation,the uneven distribution of gray information is not considered.In this paper,Otsu method is fused into the level set Chan-Vese model,and a new energy function is constructed.Reserved the advantages of Chan-Vese model and combine the information of inter class variance image distribution.So the uneven distribution of gray information is achieved.Experiment based on two datasets show that the proposed method has obvious advantages in similarity measure and error rate measurement compared with other similar methods.(2)Study on feature extraction and recognition of lesion images.Aiming at the problem that the information of medical image lesion area is complex and the effect of single texture feature classification is poor.First,this paper integrates Hough transform and invariant moments on the basis of commonly used texture features to consider the distortion and translation.And then the linearization of these nonlinear distribution fusion information,through thesupport vector machine SVM to classify it,the lesion image and the normal image recognition.Finally,the use of a hospital to provide the image texture features and shape Feature extraction,and then SVM classification experiments.The experimental results show that the accuracy of classification has improved.