Development of Low Rank Sparse Matrix Decomposition for Improving Spatial and Temporal Resolutions of MRI Medical Data

Abstract

Compressive sensing (CS) is well known method for solving the sensing, compression. Spatial and temporal resolution is one of the major limitations of dynamic Magnetic Resonance Imaging (MRI) data. These data are common in cardiac perfusion, time-resolved angiography, abdominal and breast perfusion. This paper focuses on the application of Compressed Sensing technique towards addressing the above problem. In this regard, it is essential and hence aimed to develop a sparse matrix decomposition model to efficiently represent the dynamic MRI data through enforcing sparsity. Also, the background and dynamic components of the MRI data is to be separated in a robust way than the conventional methods. The enhancement in spatial and temporal resolution is achieved as a result of high acceleration and background separation enabled by the sparse matrix decomposition model without the need of subtraction or modelling found in conventional methods. The experimentation of low rank plus sparse and Gaussian random matrix with orthogonal matching pursuit and subspace tracking algorithms are applied in cardiac MRI and Coronary Artery Disease (CAD) cardiac MRI images. The experimentation result shows that low rank plus sparse matrix with orthogonal matching pursuit algorithm performs well on medical images. The MRI data with high spatial and temporal resolutions helps medical professionals towards the diagnosis of various diseases through MRI data and on the whole it would bring benefits to the society.