BOSCO: Parallel Image Reconstruction Based on Successive Convolution Operations
Meyer C, Hu P
University of Virginia
We present a new approach for parallel imaging reconstruction for non-Cartesian trajectories: BOSCO (Based On Successive Convolution Operations). In our method, a normal gridding is performed for the data from each coil, followed by another convolution that suppresses the aliasing artifacts in one coil. The un-aliased sub-images from all the coils are combined using square-root of sum-of-squares to form the final image. In vivo spiral images show that BOSCO gives aliasing free images under acceleration factors of two and four, while maintaining sufficient SNR. As the two convolutions can be calculated relatively fast, BOSCO is a promising technique for real-time parallel imaging applications.