Enhancing Image Reconstruction via Phase-Constrained Data in an Iterative Process

This paper proposes an innovative approach to improve residual artifacts in image-based parallel magnetic resonance imaging (MRI) reconstruction. Despite its superior signal-to-noise ratio (SNR) over the conventional Sensitivity Encoding (SENSE) method, SENSE is hindered by persisting residual artifacts, causing it to be less effective in image-based parallel MRI reconstruction. We propose a joint estimation of actual and virtual coil sensitivity maps, along with the reconstructed image. Inspired by the principles of the Joint Sensitivity Encoding (JSENSE) method, the proposed approach employs an iterative optimization process via phase-constrained data of virtual conjugate coils, progressively refining these integral components to achieve superior image quality. Experimental results show that the proposed method not only enhances MRI image quality by suppressing residual artifacts but also paves the way for future research into the potential of virtual conjugate coils in image-based MRI reconstruction. Different from the phase-constrained data for enhancing k-space-based parallel MRI, the method shows that the phase-constrained data also improve image-based parallel MRI reconstruction.