Based on the observation that motion corruption and contrast pairs do not exist in a separable way in clinically obtained pediatric contrast-enhanced scans, we developed a neural network for both motion correction and optional non-contrast-enhanced image synthesis for contrast-enhanced pediatric brain MRI. We designed a neural network architecture and training schema specific to this task. We found that motion correction performance was not degraded by doing contrast synthesis simultaneously.
This abstract and the presentation materials are available to members only; a login is required.