Experiments with Content-Based Image Retrieval for Medical Images

Content-Based Image Retrieval (CBIR) is to retrieve digital images from an image data repository by the contents in the image, such as shape, texture, color, and other information that can be extracted from the image. CBIR is also referred to as query by image contents. As large image collections being created and become available, more and more applications rely on CBIR techniques to retrieve images from the collections. One important application area is the medical field. Many medical and health care institutions have started using various CBIR systems to assist and improve diagnosis and treatment of diseases. In this paper, we introduce a method for image retrieval and classification using low-level image features. This method is based on selection of prominent features in the high dimension feature space and the parameter of the k-NN algorithm. We also combine non-image features (patient records) and image features to improve the accuracy of the results. Both the patient data and images are from a clinical trail studying aloe in treating the side effects due to radiation on oral cancer patients at Mid-Michigan Medical Center. A MatLab image engine is used for image feature retrieval, and principal component analysis is applied to reduce the feature space for optimizing the performance.