Ashish Raj1, Gurmeet Singh2, Yi Wang, Ramin Zabih3
1Radiology, Weill Medical College of Cornell University, New York, NY, USA; 2Electrical Engineering, Cornell U; 3Computer Science, Cornell U, Ithaca, NY
Among recent parallel MR imaging reconstruction methods, a Bayesian method called Edge-preserving Parallel Imaging with GRAph cut Minimization (EPIGRAM) demonstrated significantly improved SNR, edge preservation and visual quality over conventional regularized SENSE method. Unfortunately, EPIGRAM requires a large number of steps in proportion to the number of intensity labels in the image, making it computationally expensive for images with high dynamic range. Here a new jump-move based graph-cut algorithm is presented that provides a logarithmic reduction in reconstruction time (typically 25-50 times) while maintaining SNR reported by EPIGRAM. Preliminary in-vivo validation for coronary angiography and short axis cine at acceleration factors of 3 and 4 are reported. Our proposal constitutes a critical step towards real-time reconstruction of cardiac MRI using the EPIGRAM approach.