Grouping of Brain MR Images via Affinity Propagation

The human brain anatomy is extremely variable across individuals in terms of its size, shape, and structure patterning. In this paper, a novel method is proposed for grouping brain MR images into different patterns. This method adopts the affinity propagation methodology to partition a population of brain images into different clusters. In the affinity propagation method, the tissue-segmented and anatomically-parcellated images are used to define the similarity between brain images, in contrast to intensity-based similarity measurement used in previous methods. After clustering, in each cluster (called a sub-group) a representative exemplar image is identified as the single subject atlas for the sub-group. Meanwhile, all the subject images belonging to the same sub-group are identified. This method has been applied to the publicly available OASIS neuroimaging dataset that includes 414 subject brain MRI images. Experiments show that the method is able to group brain MR images into different patterns effectively.