Research On Image Retrieval Based On Sparse Representation Of Liver Lesions

With the development of medical imaging equipment,medical imaging technology is playing a more and more important role in clinical application,the content of medical images is so rich that it has become an urgent problem to find the needed information from the complex medical image database.Content-based image retrieval technology plays an important role in medical diagnosis,which can assist doctors making better decisions.This paper uses sparse representation to extract the features of liver lesions CT images to realize retrieval of liver lesions CT images.The main work and innovations were as follows:(1)Discussing the theory of sparse representation,which is based on K-SVD and OMP algorithms.(2)Using sparse representation extracts the features of CT images of liver lesion.First,we extract the SIFT features of the lesion area of the liver lesion CT image and divide it into blocks,then we learn the training data set by the sparse representation method,and the learning process is divided into two stages:using K-SVD algorithm train dataset to construct over complete dictionary and using OMP algorithm to solve sparse coefficient.The sparse coefficient is the optimal feature vector which can represent the CT images of liver lesion.(3)Achieving the retrieval of CT image of liver lesions.In order to improve the performance of the retrieval,the dictionary size and sparsity of the two important parameters of the sparse coding are discussed and optimized.Experimental results confirm that when the dictionary size is set to the 32 and sparsity set to 10,the retrieval effect reaches the best.In our study,the performance of precision,recall and F1-measure were used to evaluate the performance of the retrieval.A lot of experiments were done on the data of different phases and the data of different diseases.The experimental results show that the proposed method is better than the traditional method based on gray histogram(gray histogram feature extraction of liver lesions CT image and the similarity matching)based on gray level co-occurrence matrix(extracting the texture features of liver disease CT images),can get better retrieval performance.On the other hand,the method proposed in this paper expands the application field of liver lesion CT image retrieval,and provides a new idea for studying CT image retrieval of liver diseases.The conclusion is helpful for doctors to make reasonable judgement and treatment for patients.