Improved k-t BLAST for fast fMR imaging.

ABSTRACT

A popular dynamic imaging technique, k-t BLAST (ktB) is studied here for fMR imaging. ktB utilizes correlations in k-space and time, to reconstruct the image time series with only a fraction of the data. The algorithm works by unwrapping the aliased Fourier conjugate space of k-t (y-f-space). The unwrapping process utilizes the estimate of the true y-f-space, by acquiring densely sampled low k-space data. The drawbacks of this method include separate training scan, blurred training estimates and aliased phase maps. The proposed changes are incorporation of phase information from the training map and using generalized-series-extrapolated training map. The proposed technique is compared with ktB on real fMRI data. The proposed changes allow for ktB to operate at an acceleration factor of 6. Performance is evaluated by comparing activation maps obtained using reconstructed images. An improvement of up to 10 dB is observed in the PSNR of activation maps. Besides, a 10% reduction in RMSE is obtained over the entire time series of fMRI images. Peak improvement of the proposed method over ktB is 35%, averaged over five data sets.