An approximation approach to measurement design in the reconstruction of functional MRI sequences

In Brain and Health Informatics, pp. 115-125. Springer International Publishing, 2013 (2014)

The reconstruction quality of a functional MRI sequence is not only determined by the reconstruction algorithms but also by the information obtained from measurements. This paper addresses the measurement design problem of selecting k feasible measurements such that the mutual information between the unknown image and measurements is maximized, where k is a given budget. To calculate the mutual information, we utilize correlations of adjacent functional MR images via modelling an fMRI sequence as a linear dynamical system with an identity transition matrix. Our model is based on the key observation that variations of functional MR images are sparse over time in the wavelet domain. In cases where this sparsity constraint obtains, the measurement design problem is intractable. We therefore propose an approximation approach to resolve this issue. The experimental results demonstrate that the proposed approach successes in reconstructing functional MR images with greater accuracy than by random sampling.