Breaking the noise floor

[A: magnitude signals. B: transformed signals ]

The framework for breaking the noise floor is a relatively new and unique approach for correcting bias in magnitude reconstructed MR signals.The basic idea is to transform the Rician-distributed signals (obtained from a single-receiver-coil MRI system) or the non-Central Chi signals (obtained from a N-receiver-coil MRI system with N>1) so that the transformed signals are Gaussian-distributed. This framework is applicable to MR images with either spatially uniform or nonuniform noise. In the former, the noise level (the Gaussian noise SD) can be estimated from simple sample statistics or more sophisticated techniques such as PIESNO. To read up on this aspect of MR signal and noise, you might want to visit here.

For those of you who have access to my software, below is a Matlab file detailing the steps needed to use the framework based on the one-dimensional penalized spline.

Based on a request from a student, Mr. Liu, at the University of Illinois at Urbana-Champaign, here is the Matlab file I developed for calling the two-dimensional framework based on the spherical spline. This example uses the uniform point set discussed here.