Technological breakthroughs towards motion-robust super-resolution quantitative magnetic resonance imaging for improved detection of brain diseases

URL:

Abstract:

Magnetic resonance imaging (MRI) is one of the most widely used neuroimaging techniques, thanks to its exceptional soft tissue contrast and intrinsic safety for patients. Unfortunately, signal intensities in conventional MRI images are expressed in relative units that depend on scanner hardware and software. While this poses no issue for visual inspection of anatomy, it complicates the quantitative comparison of signal intensities within a scan, between successive scans, and among different subjects. In contrast, quantitative MRI (qMRI) generates quantitative maps of absolute biophysical tissue parameters by voxel-wise fitting a biophysical signal model to a series of contrast-weighted scans. Increasing evidence suggests that qMRI can detect and quantify subtle microscopic tissue damage, aiding in the early detection and accurate diagnosis of various neurodegenerative diseases such as Alzheimer's or multiple sclerosis. The impact of these diseases is growing in our rapidly aging society. However, current qMRI techniques require long scan times due to the need for multiple contrast-weighted scans with high-resolution and high signal-to-noise ratio (SNR). Long scan times increase the risk of motion artifacts, reduce patient throughput, and decrease patient comfort, hindering the clinical adoption of qMRI. This work therefore investigates the use of model-based super-resolution reconstruction (SRR) for qMRI, aiming to optimize the trade-off between spatial resolution, SNR, and scan time. SRR enables the reconstruction of a high-resolution scan or parameter map from a series of low-resolution MRI scans, with typically shorter scan time per image. As an additional innovation in this thesis, patient movement is jointly estimated, which is crucial to avoid motion artifacts in the reconstruction. The new SRR framework is extensively tested and compared with existing SRR methods, both in computer simulations and using real brain MRI data. Furthermore, the new framework uses Bayesian statistics to incorporate prior knowledge about noise and brain tissue characteristics into the reconstruction process. This allows for the reconstruction of a 3D high-resolution quantitative tissue parameter map from motion-affected low-resolution MRI scans. Application to real brain MRI shows that the proposed framework significantly improves the accuracy of tissue parameter quantification compared to methods with a separate motion correction step. Finally, the SRR framework is extended to perfusion MRI for improved quantification of cerebral blood flow (CBF), in combination with Arterial Spin Labeling (ASL). As a result, a high-resolution quantitative CBF parameter map can be directly reconstructed from low-resolution ASL MRI scans, significantly improving SNR and CBF accuracy over traditional ASL methods.

• Google Scholar
• DOI
• BibTeX

Files: 

Download PhD Thesis