Ultra-fast (milliseconds), multi-dimensional RF pulse design with deep learning
Purpose: Some advanced RF pulses, like multi-dimensional RF pulses, are often long and require substantial computation time due to a number of constraints and requirements, sometimes hampering clinical use. However, the pulses offer opportunities of reduced-FOV imaging, regional flip-angle homogenization, and localized spectroscopy, e.g., of hyperpolarized metabolites. We propose a novel deep learning approach to ultra-fast design multi-dimensional RF pulses with intention of real-time pulse updates. Methods: The proposed neural network considers input maps of the desired excitation region of interest, and outputs a multi-dimensional RF pulse. The training library is retrieved from a large image database, and the target RF pulses trained upon are calculated with a method of choice. Results: A relatively simple neural network is enough to produce reliable 2D spatial-selective RF pulses of comparable performance to the teaching method. For binary regions of interest, the training library does not need to be vast, hence, re-establishment of the training library is not necessarily cumbersome. The predicted, single-channel pulses were tested numerically and experimentally at 3 T. Conclusion: We demonstrate a relatively effortless training of multi-dimensional RF pulses, based on non-MRI related inputs, but working in an MRI setting still. The prediction time of few milliseconds renders real-time updates of advanced RF pulses possible.