Matching algorithm with relevance feedback for brain MRI

Efficiently and accurately matching the MRI of brain in database is very important for doctors to find the similar cases. However the definition of similarity of cases may vary according to different doctors´ experience. A matching algorithm for brain MRI is presented in this paper, with the semantic label consisting of the number of regions, the size, the location and the shape feature of each region of the MRI of brain. Future more, to adapt doctors´ preferences, the relevance feedback is taking into consideration to update the weight of each feature and to save the doctors´ preferences. The testing results show that the algorithm gives a good performance.