In vivo multisite oximetry using EPR-NMR coimaging

Abstract

Coimaging employing electron paramagnetic resonance (EPR) imaging and MRI is used for rapid in vivo oximetry conducted simultaneously across multiple organs of a mouse. A recently developed hybrid EPR-NMR coimaging instrument is used for both EPR and NMR measurements. Oxygen sensitive particulate EPR probe is implanted in small localized pockets, called sites, across multiple regions of a live mouse. Three dimensional MRI is used to generate anatomic visualization, providing precise locations of implant sites. The pO₂ values, one for every site, are then estimated from EPR measurements. To account for radio frequency (RF) phase inhomogeneities inside a large resonator carrying a lossy sample, a generalization of an existing EPR data model is proposed. Utilization of known spectral lineshape, sparse distribution, and known site locations reduce the EPR data collection by more than an order of magnitude over a conventional spectral-spatial imaging, enhancing the feasibility of in vivo EPR oximetry for clinically relevant models.

Figure 1

Illustration of multisite oximetry for three probe sites. Each site covers only a small fraction of FOV. The unknowns at each site include spin densities b‾ for all voxels, a linewidth τ, and an RF phase α.

Figure 2

A composite resonator and sample holder. EPR-NMR coimaging composite resonator assembly (a). Both resonators are connected end-to-end and slide together from one end to another, bringing the stationary sample inside one of the resonators. The polyethylene tubes acts as a housing for the sample holder. Sample holder made from 50 ml centrifuge tube (b). The nose cone end of the tube is connected to the gas cylinder while the other end is connected to a plastic strip, called slider, used to slide the centrifuge tube inside the polyethylene tube.

Figure 3

Input (left) and reconstructed (right) 3D linewidth maps for one of the realizations of the eight-site digital phantom. An ROI consisting of eight 2.5 mm-radius spheres was used. The minimum distance between the sites was one voxel (2 mm). The color at each site encodes the linewidth value.

Figure 4

Curve fit for one of the projections for the digital phantom shown in Fig. 3. A projection simulated at 24.3 μT/mm gradient strength (a); dotted line indicates the nonlinear least-squares fit using the proposed forward model (b); and the residual (c).

Figure 5

Simulation results illustrating the impact of ROI size on the estimation of linewidth τ at each of the eight sites. In one case, the ROI consisted of 2.5 mm-radius spheres, while in the other case the ROI consisted of 5.0 mm-radius spheres. In both the cases, the spheres were centered around the true locations. A total of ten trials were considered.

Figure 6

MRI overlaid with low-resolution distorted EPRI spin density map. The first row shows three axial MRI slices with each passing through one of the spin deposit sites. Slices 56, 74, and 86 display back, right leg, and left leg implants, respectively. The red circles highlight the implant locations. The second row displays the respective slices from the distorted low-resolution EPRI spin density map from stage 1. Third row shows the superposition of the top two rows. The areas with no signal from both EPRI and MRI images were partially cropped resulting in a 48 × 48 mm² display.

Figure 7

Isosurface rendering of MRI and EPRI spin density map. From left to right, 3D MRI, final 3D EPRI spin density map generated from Eq. 2, and superposition of the two. For the EPRI reconstruction, the ROI was selected from the MRI.

Figure 8

MRI and multisite oximetry results superimposed. The first row shows three slices from the stage 2 EPRI spin density map. The third row shows the same three slices from the stage 2 EPRI pO₂ map. The second and fourth rows display fused EPRI-MRI for spin density and pO₂, respectively. The areas with no signal from both EPRI and MRI images were partially cropped resulting in 48 × 48 mm² display.

Figure 9

Curve fit for one of the projections collected under room air breathing. A projection measured at 24.3 μT/mm gradient strength (a); dotted line indicates the nonlinear least-squares fit using the proposed forward model (b); and the residual (c).

Figure 10

Results from in vivo multisite oximetry. The pO₂ values under carbogen breathing are compared with the pO₂ under room air breathing. Also, the results from two different ROIs, one selected from the MRI (6 mm-radius spheres) and the other selected from low-resolution EPR (9 mm-radius spheres), are compared.