Iterative & Joint Reconstruction from Calibration and Image Data for Parallel Imaging

Yu Li¹, Charles L. Dumoulin¹

The presented work introduces a new approach to reconstructing images iteratively and jointly from calibration and acquired image data in parallel imaging for accelerating MRI. Unlike a conventional parallel imaging technique that seeks a reconstruction solution "unidirectionally" from calibration to real image data, the new technique optimizes the reconstruction "bidirectionally" between low resolution/fully sampled calibration data and high resolution/undersampled real image data. This iterative and joint reconstruction offers high image quality when acquired data, including both calibration and image data, are not sufficient for a conventional reconstruction technique, thereby improving parallel imaging performance.