Lucas Lemos Franco

Thesis Topic: Deep learning reconstruction based arterial spin labelling MRI

Supervisors: Ze Wang (University of Maryland), Florian Knoll

Description

Magnetic Resonance Imaging (MRI) is an imaging modality that has significantly improved diagnostic medicine. It is a non-invasive and non-ionizing method, offering high soft tissue contrast. However, the relatively long scan time poses challenges, especially for time-sensitive techniques. By acquiring data in k-space using a non-Cartesian approach, such as a spiral trajectory, faster acquisitions are possible while preserving image quality. Additionally, the combination with Parallel Imaging techniques, which leverage redundancy in data from different coils for reconstructing missing data, enables even faster acquisitions [1].

One MRI technique that relies on fast scanning is the Arterial Spin Labeling (ASL) technique. This method focuses on visualizing blood flow by labeling the blood and waiting for it to reach the region of interest. Commonly used for assessing cerebral blood flow, ASL images originate from a very low MR signal, impacting the Signal-to-Noise Ratio (SNR) and image resolution.

Since ASL reconstruction often involves undersampled reconstruction, this thesis will explore the application of the state-of-the-art reconstruction method called End-to-End Variational Network [3] on ASL data. During model training, the training data is undersampled using a random mask. The thesis aims to investigate whether better results can be achieved for ASL by applying a spiral mask instead.

[1] Lustig, Michael. “Sparse MRI.” Stanford University, 2008. Doctoral Dissertation. Submitted to the Department of Electrical Engineering and the Committee on Graduate Studies.
[2] Alsop, D. C., et al. (2015). Recommended implementation of arterial spin-labeled perfusion MRI for clinical applications: A consensus of the ISMRM Perfusion Study Group and the European Consortium for ASL in Dementia. Magnetic Resonance in Medicine.
[3] Sriram, Anuroop, Jure Zbontar, Tullie Murrell, Aaron Defazio, C. Lawrence Zitnick, Nafissa Yakubova, Florian Knoll, and Patricia Johnson. “End-to-End Variational Networks for Accelerated MRI Reconstruction.” 2020. arXiv preprint arXiv:2004.06688.