Compressed Sensing And Joint Acquisition Techniques In Mri

The relatively long scan times in Magnetic Resonance Imaging (MRI) limits some clinical applications and the ability to collect more information in a reasonable period of time. Practically, 3D imaging requires longer acquisitions which can lead to a reduction in image quality due to motion artifacts, patient discomfort, increased costs to the healthcare system and loss of profit to the imaging center. The emphasis in reducing scan time has been to a large degree through using limited k-space data acquisition and special reconstruction techniques. Among these approaches are data extrapolation methods such as constrained reconstruction techniques, data interpolation methods such as parallel imaging, and more recently another technique known as Compressed Sensing (CS). In order to recover the image components from far fewer measurements, CS exploits the compressible nature of MR images by imposing randomness in k-space undersampling schemes. In this work, we explore some intuitive examples of CS reconstruction leading to a primitive algorithm for CS MR imaging. Then, we demonstrate the application of this algorithm to MR angiography (MRA) with the goal of reducing the scan time. Our results showed reconstructions with comparable results to the fully sampled MRA images, providing up to three times faster image acquisition via CS. The CS performance in recovery of the vessels in MRA, showed slightly shrinkage of both the width of and amplitude of the vessels in 20% undersampling scheme. The spatial location of the vessels however remained intact during CS reconstruction. Another direction we pursue is the introduction of joint acquisition for accelerated multi data point MR imaging such as multi echo or dynamic imaging. Keyhole imaging and view sharing are two techniques for accelerating dynamic acquisitions, where some k-space data is shared between neighboring acquisitions. In this work, we combine the concept of CS random sampling with keyhole imaging and view sharing techniques, in order to improve the performance of each method by itself and reduce the scan time. Finally, we demonstrate the application of this new method in multi-echo spin echo (MSE) T2 mapping and compare the results with conventional methods. Our proposed technique can potentially provide up to 2.7 times faster image acquisition. The percentage difference error maps created from T2 maps generated from images with joint acquisition and fully sampled images, have a histogram with a 5-95 percentile of less than 5% error. This technique can potentially be applied to other dynamic imaging acquisitions such as multi flip angle T1 mapping or time resolved contrast enhanced MRA