Detection Of Lesion Changes Based On Image Registration

With the rapid development of medical image detection technology,CT images are used to detect changes in lesions during the treatment of lung cancer,thereby assisting doctors in diagnosis.This article focuses on the detection of lesion changes in lung cancer CT images.First,the lung cancer CT images are registered,then the lung cancer CT images are superpixel segmented based on the image registration,and finally the lung cancer CT images are superpixel segmented on the basis of the lung cancer CT images.Lesion extraction,as well as lesion analysis.The main research contents of the paper are as follows:First,in this paper,aiming at the problem of deformation of CT images of patients before and after treatment during the detection of lesion changes,the feature-guided Gaussian mixture model was used to study the affine registration of CT images of lung cancer.In this paper,three commonly used feature extraction algorithms: Harris algorithm,SUSAN operator,and SIFT algorithm are compared and analyzed.It is found that when the brightness,rotation angle,and scale of an image change,the feature points extracted by SIFT algorithm have a good The stability.Then,based on the affine transformation model,the feature points extracted by SIFT algorithm are used to construct a Gaussian mixture model for image registration,and then the semi-supervised expectation maximization algorithm is used to solve the constructed model.Experimental results show that this algorithm can obtain more correct feature point matching relationships,thereby improving the accuracy and success rate of image registration.Secondly,in order to extract the lesions,this paper studies the method of superpixel segmentation of CT images of lung cancer based on the registration of CT images of lung cancer.This paper first builds a local Gaussian mixture model based on the traditional Gaussian mixture model.In the model parameter estimation process,eigenvalue decomposition is introduced,and the spatial position and gray value are weighed by adjusting the lower limit of the spatial covariance matrix and the gray variance.Relative importance,in order to adjust the performance of superpixel blocks in invariance and regularity.Then,combining the gray features of the lesions in the CT image of lung cancer with consistent characteristics,after superpixel segmentation of the CT image of lung cancer,outlier superpixel blocks were selected to achieve the purpose of lesion extraction.It is verified by experiments that the superpixel segmentation method in this paper can generate superpixels with controllable regularity and outperform the participating contrast superpixel segmentation methods in segmentation accuracy,and the lesion extraction method in this article is superior to several other contrast algorithms in accuracy..Finally,based on the extraction of the lesions,this article uses pixel-level measurement methods to measure the location of the lesion,and introduces the change in the number of pixels of the target lesion as an auxiliary evaluation constraint.The experimental results are analyzed.Experiments show that the analysis results after adding auxiliary constraints in this paper are basically consistent with the diagnosis results of experts;the experimental results show that the detection results of lesion changes in this article can help doctors to diagnose.