An Atlas Based Fuzzy Method for Putamen Detection in Anatomical MR Images

Recent research discloses that the putamen in human brain has close relationship with some neurological diseases, and the most commonly used methodology in such studies is magnetic resonance (MR) imaging. In order to measure the volume of putamen in MR image, accurate detection is highly desirable. In practice, both the existence of fiber in putamen and the inhomogeneity effect in MR imaging process generates gradual changes in white-to-gray matter contrast in putamen area, which makes the segmentation task rather difficult. In this paper, a Talairach atlas based adaptive fuzzy rule base (FRB) system is proposed and applied to human brain MR images for putamen detection. Comparisons to standard region growing approach, K-means algorithm, Expectation Maximization (EM) algorithm, and FMRIB Software Library (FSL)′s Automatic Segmentation Toolbox (FAST) are performed, and the proposed atlas based FRB system demonstrated significantly better performance in putamen segmentation.