UC Riverside Associate Professor of Bioengineering Jia Guo has developed a new way to reconstruct magnetic resonance imaging (MRI) scans that could improve the reliability of blood flow measurements when MRI signals are extremely weak.
The method is described in Guo’s paper, “Reference-Superimposed Reconstruction (RS-Recon) for Arterial Spin Labeling,” published in Magnetic Resonance in Medicine. The study introduces a reconstruction approach designed to improve arterial spin labeling (ASL), a noninvasive MRI technique that measures blood flow by using water in a person’s own blood as a natural tracer.
Because ASL does not require an injected contrast agent, it is used to study blood flow in the brain and other organs, and has applications in research involving stroke and other neurological disorders.
Getting the clearest ASL image, however, presents a technical challenge. MRI systems can suppress signals from background tissue so that the much smaller blood-flow signal is easier to detect. But when that background suppression becomes very strong, the remaining MRI signal can approach zero. Under those conditions, conventional reconstruction methods can flip the sign of signals, introduce artifacts, or incorrectly estimate the phase of the signal.
Guo’s method, called reference-superimposed reconstruction, or RS-recon, uses high-SNR reference data that are already routinely collected during ASL imaging. The technique temporarily combines that reference with the weaker ASL data during reconstruction, helping the system preserve the correct signal and estimate its phase more accurately. The reference information can be then removed from the reconstructed image in later processing.
In experiments involving four healthy volunteers, Guo tested RS-recon with all three forms of arterial spin labeling: pulsed ASL, pseudo-continuous ASL, and velocity-selective ASL. The method corrected artifacts caused by signal errors, recovered weak or negative signals that conventional reconstruction could lose, and improved signal-to-noise performance under several levels of background suppression.
The study also showed that RS-recon could be incorporated into existing reconstruction approaches, including GRAPPA parallel imaging, which is used to accelerate MRI acquisition. The simplest implementation tested, combining conventional magnitude reconstruction with RS-recon, produced performance comparable to more complex reconstruction methods, according to the study.
That compatibility could be important for future adoption. Instead of avoiding aggressive background suppression because of reconstruction errors, researchers could potentially suppress more unwanted tissue signals for better SNR while retaining accurate measurements of blood flow.
The work remains an early technical study. It involved a small number of healthy volunteers, and additional studies are needed to evaluate the approach across other MRI acquisition methods and in clinical populations. Guo also notes that the underlying reconstruction principle may eventually be useful for other MRI applications in which researchers need to recover extremely weak signals from scans that include a stronger reference measurement.
The research was supported by the National Institutes of Health through grants R01CA284172 and R01EB033210.
Read the study: https://doi.org/10.1002/mrm.70540.
