Breaking the Barriers of MRI by Exploiting Signals Structure
Magnetic Resonance Imaging (MRI) is an interdisciplinary field involving physics, engineering, chemistry, mathematics, and neuroscience. It is the best imaging modality for soft tissues and is considered safe as there is no exposure to ionizing radiation. However, it is still highly limited by the physics, in terms of slow scanning time, limited resolution and lack of quantitative information. Recent advances in signal processing theories have enabled breaking those barriers.
I will start the talk by presenting the physical principles of MRI and the conventional solutions to improve MRI by undersampling. I will describe several MRI applications, including structural, functional and quantitative MRI, each involves a unique structure of acquired data. I will show how the methods we developed, that rely on exploiting redundant information within and between MRI scans via sparse-based reconstruction and low-rank modeling, can significantly improve each of those applications. I will show that the theoretical bounds we developed for our methods outperform bounds developed for existing approaches and support the results obtained.