|
Abstract:
|
In fMRI, it is important to acquire volume images of the brain as rapidly as possible to capture cognitive temporal dynamics. The limiting temporal factor in fMRI is the sampling time for an array of spatial frequencies to reconstruct an image. The multi-coil SENSE in-plane reconstruction technique has greatly contributed to decreasing image scan time. However, the SENSE image reconstruction technique involves using a priori calibration images to estimate coil sensitivities, then using these estimated coil sensitivities as if they were precisely known to reconstruct images. Here, the a priori calibration image information is quantified in terms of a prior distribution on the reconstructed image and coil sensitivities, then combined with the data likelihood in order to form a posterior distribution that the reconstructed image is estimated from. This Bayesian SENSE (BSENSE) formally brings the prior image information to bear while estimating the reconstructed image in one step from the posterior distribution. BSENSE is applied to simulated data.
|
