Abstract #4400

Matthew J. Muckley ^{ 1,2 } , Douglas C. Noll ^{ 1,2 } , and Jeffrey A. Fessler ^{ 1,2 }

Algorithms that minimize SENSE-type image reconstruction
cost functions almost always compute the gradient of a
data consistency term at each iteration of the
algorithm. Incremental gradient methods approximate the
full gradient of the data consistency term by computing
the gradient using a subset of the data. Since these
subset gradients require less computation time, using
them as a proxy for the full gradient significantly
accelerates convergence. The method is general enough to
be applied to any MR image reconstruction problem
involving multiple receive coils with a SENSE model.
Four-fold acceleration is shown with a low rank plus
sparse model.

**This abstract and the presentation materials are available to members only;
a login is required.**