A method to improve time resolution in contrast enhanced magnetic resonance angiography (CE-MRA) is proposed in which a temporal basis is formed from prior knowledge of the contrast flow dynamics. The method is applicable to a number of data acquisition sequences used with multiple receiver coils to achieve a relatively high temporal resolution in post-processing.
Data is acquired in 2D or 3D k-space, sampling along one line or curved trajectory at a time. To achieve high temporal resolution, images are reconstructed from a small fraction of the data required according to the Nyquist criterion. Rather than treating each image in the sequence as independent of others, we fit a set of temporal basis functions. Our algorithm can be summarized as follows: A set of gamma variate functions is formed with different randomly chosen characteristics; the Karhunen-Loeve transform is applied to the set by computing the eigenvectors of the covariance matrix; the eigenvectors associated with the largest eigenvalues are chosen as the temporal basis functions; all pixels (voxels) are projected onto the temporal basis.
The method has been tested on simulated spiral-sampled data and on actual MR scan data from a normal subject undergoing CE-MRA. The use of the temporal basis results in a marked lowering of artifacts caused by the high degree of k-space under-sampling. Sequence rates as high as 4 frames per second have been achieved with clinically useful quality.