We propose a model-based deep learning architecture for the reconstruction of highly accelerated diffusion MRI. We introduce the use of a pre-trained denoiser as the regularizer in a model-based recovery for diffusion weighted data from k-q under-sampled acquisition in a parallel MRI setting. The denoiser is designed based on a general tissue microstructure diffusion signal model with multi-compartmental modeling. A neural network was trained in an unsupervised manner using a convolutional auto-encoder to learn the diffusion MRI signal subspace. To demonstrate the acceleration capabilities of the proposed method, we perform MRI reconstruction experiments on a simulated brain dataset.
This abstract and the presentation materials are available to members only; a login is required.