Adaptive keyhole SVD method for dynamic MRI reconstruction

Keyhole Singular Value Decomposition (KSVD) method is known as one of the most commonly used methods for reduced encoding reconstruction of dynamic Magnetic Resonance Imaging (MRI). This paper provides a new enhanced version of this method, called Adaptive KSVD (AKSVD). Instead of original method’s direct duplication of unacquired data from the reference frame, the proposed method uses a temporal model based adaptive scheme to update the unacquired data in following frames, which better captures dynamical changes. The experiment results obtained from the simulated as well as real MRI data show that the proposed AKSVD method can produce images with much lower reconstruction error.