Abstract #3659

Marinelli L, Hardy C

To accommodate the need for higher acceleration factors, coil arrays for parallel imaging applications have, in recent times, roughly doubled the number of channels each year. The resulting data flow can overwhelm reconstruction engines. We introduce and compare a class of algorithms to reduce the number of effective channels through linear combinations and elimination of coils that do not contribute significantly in the region of interest. Eigenvectors of the noise correlation matrix can be used to build a basis of effective sensitivity functions at each point in space. We validate these methods with a 32-channel cardiac array and acceleration R=1,2,4.