A flexible deep learning based framework is presented for the recovery of microstructural parameter maps from advanced diffusion models. The method is shown to work well across field strengths, and healthy and diseased tissue without needing separate training of the DL network. The DL framework is embedded in a model-based reconstruction, which enables the framework to handle variations in data acquisition settings such as various acceleration factors and noise levels, without having to change the network. k-q accelerations in the range of 12-18 fold is demonstrated for single and multi-shell diffusion data.
This abstract and the presentation materials are available to members only; a login is required.