Study On Compressed Sensing-based Auxiliary Diagnosis Algorithm Of Ulcerative Colitis

Ulcerative colitis is a common gastrointestinal disease in adults.It recurs during the course of the disease and significantly increases the risk of colon cancer infection.The routine and effective test method is colonoscopy.However,due to the large number of endoscopic images and Contains noise such as bubbles and light spots,which brings many difficulties to the doctor's diagnosis.And the existing data labeled by physicians is relatively scarce,which also limits the development of deep learning applications in this scenario.Therefore,designing an auxiliary diagnostic system for ulcerative colitis is a valuable and meaningful study.This article studies the process of finding the sparse coefficient of compressed sensing,and finds that it is similar to the feature representation of the sparse feature model for small sample classification problems.The compressed sensing system is introduced into the image classification model to build a new compressed sensing-based ulcerative colitis auxiliary diagnosis system Briefly introduce the main work of this article:1.An alternate optimization algorithm based on measurement matrix and sparse dictionary minimization of sparse representation error is proposed.First,block recursive least squares dictionary learning is used as the initial dictionary,and prior knowledge is used to solve the analytical solution of the initial measurement matrix.After alternate optimization,Orthogonal matching pursuit algorithm is used for image reconstruction.The peak signal-to-noise ratio of the algorithm on the actual image is improved by about 4d B compared with the conventional method.2.Compressed sensing spatial pyramid pooling image features are proposed for image-assisted diagnosis of ulcerative colitis.First,the initial dictionary and observation matrix are solved based on the prior knowledge of the intestinal image,and then the observation matrix and the sparse dictionary are optimized alternately.Under this compressed sensing framework,more accurate sparse features of the intestinal image are obtained,and finally combined with the spatial pyramid pooling characteristics,the local and global information of the data set is extracted at different resolutions,and the sparse representation coefficient is used to better distinguish the diseased image from the normal intestinal image.This paper verifies on the actual ulcerative colitis data set that the accuracy of the model is 12.35% higher than the bag of features model,3.99% higher than the sparse coding space pyramid matching model,and the accuracy is 2.27% higher than the locality-constrained linear coding model.