Fast dynamic MRI using linear dynamical system model

Abstract

Imaging of physiological functions using magnetic resonance imaging is limited due its slow data acquisition speed. Previously various techniques based on data sharing in the spatiotemporal k-space or sparse recovery methods have been proposed to increase imaging speeds in dynamic MRI. This paper presents a novel formulation for fast dynamic MRI which combines the generic linear dynamical system based spatiotemporal model with sparse recovery techniques. Specifically, the formulation uses a known prior time-evolution model for the physiological function implicitly and enforces the model errors (innovations) to be sparse. The preliminary results of dynamic MRI recovery experiments on an in-vivo myocardial perfusion dataset show that the proposed approach preserves details like edges and fine structures in recovered images better than previous k-space data-sharing and sparse recovery techniques individually.