Research On Medical Image Retrieval Based On Texture And Shape

In clinical medicine, The medical image is an important objective evidence for doctors in clinical diagnosis, disease tracking, surgical planning, prognosis researching and so on. With the continues development and completion of the electronic medical records and Picture Archiving and Communication Systems (PACS) technology, a mass of clinical image data will be producted every day. The traditional medical image retrieval technique Text-based Medical Image Retrieval have not meet the real needs. Content-based methods were adopted in medical image retrieval.The development, research status and key technology of the medical image retrieval were discussed in this paper firstly, Then the current existing medical image segmentation and feature extraction techniques were classified and analyzed. The main research work and innovation of this paper are given as:(1) Texture and shape feature extraction techniques and the medical image segmentation techniques are researched and summarized.(2) The medical image retrieval based on ellipse partition is proposed. After the ellipse partition, GLCM is adopted for texture description, and the energy, contrast, correlation, and entropy extrated from GLCM are used as texture features. Then the average of four texture features in different region are computed respectively as the final feature vector. Finally, Euclidean distance is introduced as similar measure distances.(3) Based on the study of the local binary pattern, the center of the local binary pattern (CLBP) texture feature extraction methods, a new method based on the integration of corner and texture features was proposed. Corners of the medical image are extracted based on multi-scale curvarture polynomical for shape description and CLBP are extracted for texture description. The experimental results show that, the new method can reflect the spatial information, and it can effectively improve medical image retrieval precision and recall.