Haris Saybasili1,2, Peter Kellman1, Mark A. Griswold3, J. Andrew Derbyshire1, Michael A. Guttman1
1NIH/NHLBI, Bethesda, MD, USA; 2Biomedical Engineering Institute, Bogazici University, Istanbul, Turkey; 3Case Western Reserve University, USA
We present a new parallel imaging algorithm based on TGRAPPA for real-time MRI, called HTGRAPPA and its real-time, low-latency implementation suitable for interventional MR applications. Our method calculates GRAPPA coefficients in the k-space, and transfers them in the image domain. These image domain GRAPPA weights were combined into composite unmixing coefficients using adaptive B1-weight estimates and optimal noise weighting. That way, convolution operations in the k-space are avoided during the reconstruction, thus blazingly fast reconstruction speeds are achieved. More than 70 frames per second reconstruction performance is achieved on 30 coil, rate 4 dataset (up to 265x faster than TGRAPPA).