We hypothesise that machine learning approaches could be applied to speed up motion-correction navigators – potentially obviating the need for image reconstruction and co-registration. In this preliminary 2D simulation study we investigate the ability of a simple 5-layer convolutional neural network to predict motion-parameters based on difference images between two head poses. Our results indicate that the CNN is able to outperform linear regression over the range of parameters tested, supporting our aim to develop this concept in more detailed future work.
This abstract and the presentation materials are available to members only; a login is required.