03 Sep 2019
Washington University in St. Louis and University of Birmingham use phase data to boost image quality.
A project from the University of Birmingham and Washington University in St. Louis has now demonstrated what it describes as "critical improvements to functional near-IR spectroscopy (fNIRS)-based optical imaging in the brain." The work was published in Neurophotonics.
The advance hinges on enhancements to reconstructed image quality that become possible when phase shift measurements, which reflect the time-of-flight of the NIR light, are incorporated within the tomographic reconstruction.
In previous studies using frequency-domain near-IR spectroscopy to recover hemoglobin concentrations, the phase measurements have typically been disregarded, according to the project team.
The research is immediately relevant for diffuse optical tomography (DOT), an established technique which arranges the NIR optical source and detector elements in a dense grid, providing overlapping measurements that can be turned into tomographic reconstructions through computational analysis.
Successfully applying a frequency-domain (FD) multidistance approach to high definition DOT has proven challenging in the past, due to the dynamic range needed. But the new research compared imaging resolution and accuracy between continuous wave (CW) and FD high-resolution DOT, as a way to determine if the technique could be used to achieve enhanced resolution and imaging depth in regions of the brain. It is said to be the first comparison between frequency-domain and continuous-wave approaches for functional brain imaging via DOT.
Deeper regions of the brain
"HD-DOT systems incorporating multidistance overlapping continuous wave measurements only recover differential intensity attenuation," commented the researchers in the published paper. "We investigate the potential improvement in reconstructed image quality due to the additional incorporation of phase shift measurements, which reflect the time-of-flight of the measured NIR light, within the tomographic reconstruction from high-density measurements."
The approach was tested first on models of the brain based on specific patients, and then on a modelled system with realistic levels of noise to represent real-world tissue conditions. An in vivo trial was subsequently carried out, using a FD NIRS platform of 32 sources modulated at 140 MHz, and 30 detector fibers with each source coupled to laser diodes emitting at 690 and 830 nanometers.
"To evaluate image reconstruction with and without the additional phase information, we simulated point spread functions across a whole-scalp field of view in 24 subject-specific anatomical models using an experimentally derived noise model," said the team in its paper. Results showed that addition of phase information reduced localization error by up to 59 percent, and improved effective resolution by up to 21 percent as compared to using the intensity attenuation measurements alone.
The phase data also enabled images to be resolved at deeper brain regions, as deep as 20 to 25 millimeters, where measurements based on intensity data and CW modalities completely fail.
"This study suggests that the use of frequency-domain measurements in the context of high-definition fDOT expands the potential for improvements in image quality beyond the current state of the art," concluded the researchers.