We present a data-driven technique for denoising 3D Magnetic Resonance Spectroscopic Imaging (MRSI) data. Our proposed method involves a novel spectral de-phasing and re-phasing approach which increases the phase dimension of the spectra to deal with the arbitrary complex phase in the data. This is coupled with an anisotropic non local means (NLM) filter-based pattern-recognition across the multi-slice data to select similar spectra patches having a similar phase for denoising. We show that our method leads to a mean SNR improvement by an average factor of 4.5 while preserving the spectral resolution of the metabolites.
This abstract and the presentation materials are available to members only; a login is required.