Project 1 of this consortium, Brain Data Correlation, will focus on the experimental and statistical operations necessary to understand the relationships between different structural or functional brain data sets. We will develop and evaluate a variety of approaches for mapping anatomic and physiologic data into a common probabilistic reference space. The proposed work will extend the atlas concept by providing a more adaptable and useful way of measuring and comparing neuroanatomic and neurophysiologic data. Rather than depend on a single representative data set, we will develop a probabilistic model that represents a population. There are six specific aims. First, we will acquire the appropriate data sets for the development and testing of algorithms for correlation and deformation, including phantoms, in vivo and postmortem primate anatomy and human MRI-PET with skull-based fiducials. Second, we will develop corrective schemes to minimize errors and variance introduced by the acquisition process itself. Third, reliable and accurate algorithms (positioning, scaling, affine transformations, linear and non-linear transformations) for placing data volumes within reference/coordinate systems will be defined. Approaches designed for one modality may not be applicable for others. Fourth, we will develop transformations that include local deformations to increase the degree of correlation between data sets. We will derive the procedures and mathematics to warp data sets to map one upon another, to map to a common probabilistic reference system and to retain the direction and degree of deformation in a format suitable for interacting with the probabilistic space. Fifth, we will develop approaches that will include the retention of variability. The focus of these efforts is to make more comprehensive the representation of neuroscientific information about structure and function as opposed to the traditional approach of using a single subject representation from which to base an atlas/reference system. In this way we plan to retain information about inherent brain variation that will help us understand its normal distribution. Sixth, software design and validity will be driven by our goals to quantify and optimize the tools for selecting the most ideal approach to acquisition, acquisition correction, alignment/registration, deformation and retention of variability.
Showing the most recent 10 out of 36 publications
