Thanks to powerful clinical applications, magnetic resonance (MR) parameter mapping has received widespread attentions. This work shows that the physical decay about parameter maps can be implicitly absorbed into an annihilation relation from k-space measurement of weighted MR images. Routinely, this relation can be estimated via null-space decomposition of a structured matrix, but, which usually results in computational burden. To alleviate it, we propose to train a convolutional neural network (CNN) to estimate this annihilation relation from undersampled measurement to further realize k-space interpolation. Experiments reveal the effectiveness of the proposed method compared with other competing methods.
This abstract and the presentation materials are available to members only; a login is required.