Spatially uniform sampling in 4-D EPR spectral-spatial imaging

Abstract

Four-dimensional EPR imaging involves a computationally intensive inversion of the sampled Radon transform. Conventionally, N-dimensional reconstructions have been carried out with N−1 stages of 2-D backprojection to exploit a dimension-dependent reduction in execution time. The huge data size of 4-D EPR imaging demands the use of a 3-stage reconstruction each consisting of 2-D backprojections. This gives three orders of magnitude reduction in computation relative to a single stage 4-D filtered backprojection. The multi-stage reconstruction, however, requires a uniform angular sampling that yields an inefficient distribution of gradient directions. We introduce a solution that involves acquisition of projections uniformly distributed in solid angle and reconstructs in three 2-D stages with the spatial uniform solid angle data set converted to uniform linear angular projections using 2-D interpolation. Images were taken from the two sampling schemes to compare the spatial resolution and the line width resolution. The degradation in the image quality due to the additional interpolation was small, and we achieved ∼30% reduction in data acquisition time.