Improved parallel MR imaging with accurate coil sensitivity estimation using iterative adaptive support

Abstract

Estimation of sensitivity profile in SENSE-like methods plays a crucial role in reconstruction. The self-calibration of sensitivity eliminates the separate calibrating scan and therefore reduces imaging time. However, sensitivity estimation by zero-filling in each column outside region of the object introduces inaccuracy and artifacts into the results, especially for the image periphery. Noise and error may propagate to reconstruction. In this paper, based on the method of joint sensitivity estimation and image reconstruction, penalty theory was used to reformulate the objective function to refine the sensitivity maps in each coil. The proposed method was tested on various data sets and in vivo brain data were shown for comparison. By suppressing the background and enhancing sensitivity maps in the region of interest through iterations, the quality of reconstructed image improved significantly, especially when a rather large reduction factor was used.