Felix A. Breuer1, Andre Fischer1, Nicole Seiberlich2, Philipp Ehses1, Martin Blaimer1, Daniel Neumann1, Peter M. Jakob1,3, Mark A. Griswold2
1Research Center Magnetic Resonance Bavaria, Wrzburg, Germany; 2Radiology, Case Western Reserve University, Cleveland, OH, United States; 3Experimental Physics 5, University of Wrzburg, Wrzburg, Germany
In this work we demonstrate that the concept of Principal Component Analysis (PCA) can significantly improve Compressed Sensing (CS) reconstructions of highly undersampled contrast enhanced MR Angiography (MRA) data. In contrast to conventional CS, in this approach, each CS step operates in a heavily compressed basis. After PCA the dynamics are modeled within a few PCs exhibiting very high SNR, resulting in more accurate CS reconstruction results. In addition, the new method employs an iterative update of the principal components (PCs) after each CS step and thus is self-calibrating and does not require any prior knowledge about the contrast dynamics.