Compressed Sensing for Dynamic Imaging
Compressed Sensing (CS) or sparse data sampling as applied to Magnetic Resonance Imaging (MRI) is a way of reducing the amount of raw (so called k-space) data acquired by undersampling and re-constructing the images (and quantitative maps) without artefact or distortion. The benefits are reduction in scan time in any imaging where dynamic processes are being observed (such as cardiac imaging) or quantitative imaging (e.g. relaxation mapping). Whilst these techniques have been on the fringes of use in MRI for a while, it is now clear that with the advent of fast processing hardware available and suitable non-Fourier reconstruction methods, CS is becoming a reality. The full advantage of CS techniques only become apparent in high-dimensional data sets i.e. 3D spatial and a one temporal or parameter dimension. This project would aim to implement CS algorithms for image reconstruction and implement their use in model based quantitative mapping (such as for multi-TI Arterial Spin Labelling and MR relaxometry) on the Philips 7T and 3T MRI scanners at SPMIC. These quantitative mapping methods are widely used for brain imaging at 7T (for example in functional MRI and clinical studies such as Multiple Sclerosis), as well as for body imaging at 3T (for example for renal imaging in Chronic Kidney Disease).
The position would suit a physicist with a strong interest in numerical and mathematical methods including signals and imaging processing techniques.
Dynamic Arterial Spin Labelling perfusion images taken at multi-TI values as indicated. CS will incorporate the dynamic model, allowing for more precise determination of cerebral blood flow and volume at higher resolution with improved SNR, reduced estimation error and suppressed motion artefacts.