We developed a simple deep learning method for DWI data denoising and tested it on correcting sum of square (SoS) noise. By acquiring two sets of diffusion images reconstructed with SoS and SENSE1 coil combination schemes on one subject as training data, the learned model can effectively denoise any SoS data acquired with the same DWI protocol. The denoised data produces similar results in diffusion tensor analysis and NODDI analysis as the SENSE1 data. This method also shed light on denoising techniques for diffusion imaging if a low-noise DWI dataset is available.
This abstract and the presentation materials are available to members only; a login is required.