Unbiased Diffeomorphic Shape and Intensity Atlas Creation: Application to Canine Brain

B. Avants, G. Aguirre, J. Walker, J. C. Gee University of Pennsylvania, Philadelphia, PA, United States Introduction: High dimensional deformable image registration [1] is important for spatially locating functional activation [2], understanding disease progress [3] and for improving the statistical power of clinical trials [3]. The coordinate system chosen for a particular structural or functional study influences the quantitative outcome, as the reference frame affects the measurement in the nonlinear world of morphometry [4]. The statistically conservative approach thus requires least biased coordinate systems where average anatomical configurations are estimated from a spatially normalized population [1,4,5]. Prior work on this problem falls into two categories: deformation-based averaging [1] and intensity-based averaging [5]. The advantage of intensity averaging is that it removes dependence on the intensity signature of any single, specific anatomy. Intensity averaging may create false structures by averaging tissues that are not in correspondence [5]. Another disadvantage is that the residual of the geometric (or shape) component of the average may be far from zero. Shape-based averaging, on the other hand, guarantees that tissues are in correspondence before averaging and gives a minimal shape residual. However, initialization is with respect to a specific anatomical space [1], inducing a dependence on the initial anatomy. This paper outlines a procedure that uses intensity averaging to gain an unbiased initialization for shape-based averaging, giving a best-of-both-worlds result. The method, here, provides an average dog brain atlas. The atlas defines a common canine-stereotactic space used to interpret patterns of cortical activation to visual stimuli and for identifying primary and extrastriate cortical areas.