Model-based super-resolution reconstruction with joint motion estimation for improved quantitative MRI parameter mapping

Quinten Beirinckx; Ben Jeurissen; Michele Nicastro; Dirk H J Poot; Marleen Verhoye; Arnold J den Dekker; Jan Sijbers

doi:10.1016/j.compmedimag.2022.102071

Model-based super-resolution reconstruction with joint motion estimation for improved quantitative MRI parameter mapping

Comput Med Imaging Graph. 2022 Sep:100:102071. doi: 10.1016/j.compmedimag.2022.102071. Epub 2022 May 10.

Authors

Quinten Beirinckx¹, Ben Jeurissen², Michele Nicastro², Dirk H J Poot³, Marleen Verhoye⁴, Arnold J den Dekker², Jan Sijbers²

Affiliations

¹ imec-Vision Lab, Department of Physics, University of Antwerp, Antwerp, Belgium; μNEURO Research Centre of Excellence, University of Antwerp, Antwerp, Belgium. Electronic address: quinten.beirinckx@uantwerpen.be.
² imec-Vision Lab, Department of Physics, University of Antwerp, Antwerp, Belgium; μNEURO Research Centre of Excellence, University of Antwerp, Antwerp, Belgium.
³ Biomedical Imaging Group Rotterdam, Department of Radiology and Nuclear Medicine, Erasmus MC, Rotterdam, The Netherlands.
⁴ Bio-Imaging Lab, Department of Biomedical Sciences, University of Antwerp, Antwerp, Belgium; μNEURO Research Centre of Excellence, University of Antwerp, Antwerp, Belgium.

PMID: 36027768
DOI: 10.1016/j.compmedimag.2022.102071

Abstract

Quantitative Magnetic Resonance (MR) imaging provides reproducible measurements of biophysical parameters, and has become an essential tool in clinical MR studies. Unfortunately, 3D isotropic high resolution (HR) parameter mapping is hardly feasible in clinical practice due to prohibitively long acquisition times. Moreover, accurate and precise estimation of quantitative parameters is complicated by inevitable subject motion, the risk of which increases with scanning time. In this paper, we present a model-based super-resolution reconstruction (SRR) method that jointly estimates HR quantitative parameter maps and inter-image motion parameters from a set of 2D multi-slice contrast-weighted images with a low through-plane resolution. The method uses a Bayesian approach, which allows to optimally exploit prior knowledge of the tissue and noise statistics. To demonstrate its potential, the proposed SRR method is evaluated for a T1 and T2 quantitative mapping protocol. Furthermore, the method's performance in terms of precision, accuracy, and spatial resolution is evaluated using simulated as well as real brain imaging experiments. Results show that our proposed fully flexible, quantitative SRR framework with integrated motion estimation outperforms state-of-the-art SRR methods for quantitative MRI.

Keywords: Bayesian estimation; Model-based reconstruction; Motion correction; Quantitative magnetic resonance imaging; Super-resolution.

Publication types

Research Support, Non-U.S. Gov't

MeSH terms

Bayes Theorem
Brain / diagnostic imaging
Image Processing, Computer-Assisted
Imaging, Three-Dimensional* / methods
Magnetic Resonance Imaging* / methods
Motion