Absolute Oxygen R_1e Imaging In Vivo with Pulse Electron Paramagnetic Resonance

Abstract

Purpose

Tissue oxygen (O₂) levels are among the most important and most quantifiable stimuli to which cells and tissues respond through inducible signaling pathways. Tumor O₂ levels are major determinants of the response to cancer therapy. Developing more accurate measurements and images of tissue O₂ partial pressure (pO₂), assumes enormous practical, biological, and medical importance.

Methods

We present a fundamentally new technique to image pO₂ in tumors and tissues with pulse electron paramagnetic resonance (EPR) imaging enabled by an injected, nontoxic, triaryl methyl (trityl) spin probe whose unpaired electron’s slow relaxation rates report the tissue pO₂. Heretofore, virtually all in vivo EPR O₂ imaging measures pO₂ with the transverse electron spin relaxation rate, R_2e, which is susceptible to the self-relaxation confounding O₂ sensitivity.

Results

We found that the trityl electron longitudinal relaxation rate, R_1e, is an order of magnitude less sensitive to confounding self-relaxation. R_1e imaging has greater accuracy and brings EPR O₂ images to an absolute pO₂ image, within uncertainties.

Conclusion

R_1e imaging more accurately determines oxygenation of cancer and normal tissue in animal models than has been available. It will enable enhanced, rapid, noninvasive O₂ images for understanding oxygen biology and the relationship of oxygenation patterns to therapy outcome in living animal systems.

Introduction

Importance of EPR O₂ Images

Molecular oxygen, O₂, is a crucial molecular determinant of states of human health and disease. One third of human deaths are due to diseases of O₂ deprivation such as ischemic heart disease and cerebrovascular disease (1). Regions of low O₂, hypoxia, thought to be a universal characteristic of solid tumors, (2) reduce the efficacy of radiation therapy in their treatment (3). The importance of organismal response to hypoxia is reflected in the hundreds of genes regulated by hypoxia inducible factor-1 (HIF-1). (4,5) As will be seen in animal model tumor images presented in this paper, large pO₂ gradients separate regions of high and low pO₂ throughout a tumor. In vivo quantification of tissue pO₂ in disease states, thus, requires imaging to fully characterize the oxygenation state of the entire tumor. EPR O₂ images may define the mechanism by which hypoxia creates resistance to radiation, which remains elusive. (2) pO₂ EPR images registered with either stereotactic biopsies or gene induction images can define quantitative gene induction response to local pO₂ - in vivo - in states of health and disease (6).

R₁ images solve confounding self-relaxation of O₂ EPR Images

Heretofore, in in vivo EPR O₂ images, pO₂ has been measured via the broadening of the EPR spectral lines of an injected spin probe predominantly through Heisenberg spin exchange (HSE) with local O₂. (7) The spin probe is a paramagnetic trityl molecule (8) with a single narrow EPR line whose relaxation times are long enough to enable pulse imaging. Spin packet spectral line width (LW) broadening in a continuous wave (CW) experiment is physically equivalent to increasing the transverse magnetization relaxation rate (R_2e) in a pulse experiment. LW=(γ_eT_2e)⁻¹=R_2e/γ_e; γ_e ≡ gyromagnetic ratio of the electron (9). Both R_2e and R_1e depend linearly on pO₂ (8,10,11). The spin probe collision rate with extremely rapidly relaxing O₂, increases R_2e, destroying the magnetization phase coherence. EPR images of R_2e yielded quantitative O₂ images in vivo (12–15) providing about 1 mm resolution, and a method less susceptible to confounding biologic variation than other techniques (2).

HSE between the unpaired electrons of O₂ and spin probe similarly increases the spin probe electron spin-lattice or longitudinal relaxation rate R_1e, which measures the loss of the trityl spin magnetization energy to the lattice, predominantly to the O₂ (7). However, HSE self-relaxation processes, HSE interactions between spin probes, affect R_1e and R_2e differently. HSE between two trityl molecules with differing EPR frequencies produces additional precession phase shifts that increase the R_2e (7). HSE does not alter the total energy of an interacting spin probe pair, and therefore does not affect the electron spin system energy - the longitudinal magnetization. Use of image pulse sequences dependent on R_1e reduces probe self-relaxation by nearly an order of magnitude. At this level, essentially the only physiologic R_1e relaxation rate increase is due to O₂.

Imaging of trityl concentration is difficult

Trityl concentration, referred to as [trityl], is related to the amplitude of the trityl signal in each voxel and might be useful to correct the effect of [trityl] self-relaxation on R_2e. However the accuracy in determining [trityl], using local signal amplitude is complicated by the extracellular fluid compartment distribution of trityl (16) and the variability of the extracellular volume fraction in the chaotic tumor environment. Trityl distribution volume varies from voxel to voxel making [trityl] correction difficult, blurring R_2e-based pO₂ accuracy. We find, here, R_1e pO₂ images are far less [trityl] dependent and more accurate than R_2e -based images.

Imaging longitudinal relaxation rate R_1e is straightforward and accurate with pulse EPR

In saline solutions and very low [trityl], R_1e is close to R_2e at physiologic temperatures. Although quantitative oxygen R_1e EPR measurements (11,17) and images (18) have been reported these were continuous wave studies. Pulse imaging is more efficient, shows better precision and, as we will show in this paper, accuracy. (19) This work is the first to demonstrate direct, quantitative, pulse in vivo R_1e images. The long electron relaxation times of the trityl OX063 (8) with T_1e and T_2e of ~ 7 µs at hypoxic conditions, and consequent slow relaxation rates enable pulse R_1e images with current RF technology. We read out R_1e from each voxel using inversion recovery (IR) (20), imaging the recovering longitudinal magnetization as a function of time after the inversion pulse using electron spin echo (ESE) imaging (14,21) - a tomographic frequency-encoding method. We thus refer to inversion recovery with ESE readout as IRESE.

The inversion recovery sequence IRESE (Fig. 1), inverts the spin polarization with a π–pulse preceding the ESE π/2–π detection sequence. The recovery is measured as a function of the delay T (Fig. 2) between the inversion pulse and the ESE detection sequence. The coefficient A in the equation in Fig. 2 is less than or equal to, but near, two. Values less than 2 account for incomplete EPR line inversion. A single echo time τ is fixed at an optimum slightly longer than the imager dead time.

Figure 1.

IRESE pulse sequence for R_1e imaging.

Figure 2.

Simulation of a single voxel’s signal amplitude dependence on T in IRESE sequence. T_1e = 5 µs was used.

This study compares R_1e with R_2e imaging using a 250 MHz pulse imager (14). We demonstrate that in vivo R_1e imaging shows much higher absolute pO₂ accuracy, overcoming the limitations imposed by [trityl] relaxation on R_2e imaging.

Methods

Imaging Methods

A pulse 250 MHz imager (14) was enhanced with a passive transmit-receive switch (22) and π/2- and π- pulses of equal duration/bandwidth (23). The imager was controlled with SpecMan4EPR v. 1.1.6 (24).

Image acquisition time and spatial resolution were fixed at, respectively, 10 minutes and 1.5 mm. Spatial resolution of the ESE and IRESE images were defined by the Rayleigh resolution criterion (14,25). We used the same pulse sequences for phantoms and for animal imaging. The standard deviation of the relaxation times/rates in homogeneous phantoms, excluding two outer layers to avoid partial volume artifacts, was used as an estimation of relaxation time/rate errors. Tables 1 and 2 present parameters of the R_1e and R_2e sequences. The repetition time, T_R, for ESE sequences was adjusted to keep constant the delay between the last pulse in a sequence and the first pulse of the next sequence. For accurate R_1e in IRESE images the image recorded at infinite recovery time T was equated to an image recorded without an inversion pulse. For voxel intensity fitting to an exponential recovery function, a T equal to 36 µs was assigned to this image.

Table 1.

Pulse sequences and imaging protocols.

Protocol	Description
Non-imaging two-pulse ESE	π/2-τ-π-τ-echo; 35 ns π/2 and π RF pulses; 80 τ's logarithmically spaced between 630 ns and 14 µs; 16-step phase cycling; 70 µs repetition time; echo is integrated; τs are measured in random order.
Non-imaging IRESE	π-T-π/2-τ-π-τ-echo; 35 ns π/2 and π RF pulses; τ = 630 ns; 16-step phase cycling for detection sequence; 80 Ts are spaced logarithmically between 0.5 µs and 32 µs; 80 µs repetition time; echo is integrated; Ts are measured in random order.
Two-pulse ESE imaging	π/2-τ-π-τ-echo; 35 ns π/2 and π RF pulses; time trace 1500 points with 4 ns dwell time; 16-step phase cycling, 37472 acquisitions per τ, including phase cycling; 5 τs are spaced logarithmically between 0.63 µs and 2.4 µs; TLFR = 10.37 µs; |G|=15 mT/m; imaging time 10 minutes.
IRESE imaging	π-T-π/2-τ-π-τ-echo; 35 ns π/2 and π RF pulses; time trace 1500 points with 4 ns dwell time; τ = 630 ns; 16-step phase cycling applied only for detection sequence, 9600 acquisitions per T, including phase cycling; 8 Ts are spaced logarithmically between 0.41 µs and 14 µs; TLFR = 25 µs; |G|=15 mT/m; imaging time 10 minutes.

Table 2.

Parameters of pulse sequences

	Pulse length	RF power [W]	Bandwidth [MHz]	Transmitted Average Power [W]
2pESE (T2e)	35 ns, π/2 and π	39.6 (π/2), 158.5 (π)	8.7	0.57
IRESE (T1e)	35 ns, π/2 and π	39.6 (π/2), 158.5 (π)	8.7	0.42

For images, (14,26) 208 equal solid angle projections (27) were acquired; gradient |G⃗|=15 mT/m; field of view =4.24 cm. A 53 baseline projections obtained with 1.5 mT lower main field were acquired with every fourth projection and subtracted from the previous four projections. Projection numbers were expanded with four-fold cubic B-spline interpolation (27) and were filtered with a 3D Ram-Lak filter cutoff at one half the Nyquist frequency. Voxels with amplitude less than 15% maximum at the shortest delay were eliminated (thresholded). (14).

The system frequency band-pass function for each acquisition technique was measured using zero gradient sample signal amplitude at 50 spanning B₀ fields (14). Projections were normalized using this function. Image analysis was performed with in-house software written using MATLAB (Mathworks, Inc., Natick, MA).

The spatial resolution of an image can be quantified by the response of an image to an abrupt step function change in sample density fitted with the Gauss error function (erf(x/2σ)). The width of this error function σ=1.5mm is an estimate of the ESE image spatial resolution. IRESE image has the same spatial resolution.

Spin Probe

Spin probe was the trityl OX063 radical methyl-tris[8-carboxy-2,2,6,6-tetrakis[2-hydroxyethyl]benzo[1,2-d:4,5-d’]bis[1,3]dithiol-4-yl]-trisodium salt, GE Healthcare (LittleChalfont, Buckinghamshire, UK). Phantoms with 1 mM spin probe in saline were glass cylinders 9.5 mm i.d., 45 mm long. The 0% O₂ sample was degassed via freeze-pump-thaw; the 9.3% O₂ sample was bubbled with a 9.3%:90.7% oxygen-nitrogen mixture and epoxy sealed. For measuring oxygen concentration dependence, the phantoms were bubbled in situ with humidified gas mixture. 30 minutes or more were given for oxygen equilibration during these measurements.

Non-imaging vs. imaging conditions

Acquisition of spatial information requires considerable time, which, for in vivo imaging, is limited by the animal physiology. Therefore imaging protocols have to balance the precision of the measurements with the experiment’s duration. As a result, the relaxation times in imaging protocols are estimated from only five points (R_2e) or eight points (R_1e) on the decay curve. Such restrictions do not apply for non-imaging measurements on phantoms, which have larger numbers of delays (T) from the inversion pulse (80 vs 5 R_2e – 8 R_1e for imaging) and wider T_R interval (at least 5 hypoxic T_1e for non-imaging vs ~ T_1e for imaging). Non-imaging measurements were repeated ~50 times, 20 seconds for each measurement; data were fitted independently, and the average was presented.

Animal Imaging

FSa fibrosarcomas were grown on the legs of 6–8-week-old C3H mice (HSD, Indianapolis, IN) immobilized with a partial circumference vinyl polysiloxane cast (GC Dental Products, Kasugai, Japan) (28). For Fig. 4, OX063 was injected IV 0.56 mmol/kg followed by infusion at 0.63 mmol/kg/hr. In Fig. 5, after each R_1e/R_2e image pair additional 0.21 mmol/kg OX063 was injected, and infusion was increased by 0.35 mmol/kg /hour, to a maximum of 3.85 mmol/kg /hour. This was performed on three animals with consistent results. Tumor was defined by T₂ enhancement in RARE MRI registered with EPR images (29).

Figure 4.

a–c: sagittal slice (0.7 mm) of a mouse leg bearing a tumor. a. R_2e image from ESE; b. R_1e image from IRESE; and c. [trityl] image from ESE. The tumor contour is obtained from a registered MRI. d–f: stacked histograms of d. R_2e; e. R_1e; and f. the difference between R_1e and R_2e.

Figure 5.

R_1e (◦) and R_2e (•) in a mouse vs [trityl] as spin probe is infused at different rates. [trityl] is obtained by normalization of animal signal intensity in each voxel on the voxel intensity of phantom with 1mM concentration. A sagittal slice of R_1e image of the tumor-bearing leg is shown on the left. Leg profile and tumor contour are obtained from registered T₂-weighted MRI. R_1e and R_2e [trityl] dependences for two areas [1] and [2] obtained by averaging voxels in ~8 mm³ cube (27 voxels) are shown. Error bars are the R_1e and R_2e standard deviations. Area 1 is located in the tumor, and area 2 is in the muscle. The slopes of [trityl] dependence and 50% confidence intervals are given in the plots.

Animal experiments followed USPHS policy, and were approved by the Institutional Animal Care and Use Committee.

Results

In vitro studies show reduction of R_1e sensitivity to confounding [trityl] relaxation by nearly an order of magnitude relative to that of R_2e

We performed studies in vitro by using the OX063 spin probe, as described in Methods (15,30,31). Sodium chloride concentration affects the relaxation times (8). Thus, normal saline (145 mM NaCl) resembling animal condition, was used as solvent (1). Temperature, was kept at 37°C, as in rodents and humans (32). These conditions affected the relaxation times and made them substantially different from the measurements on the same spin probe dissolved in water and performed at room temperature; see for example (33). In saline sodium ions allow the tri-acid OX063, whose charge is −3 at physiologic pH, to approach more closely to each other than in water, increasing [trityl] dependent transverse self-relaxation. The dependence of OX063 relaxation rates on pO₂ is given in Figure 3a. The experimental data were corrected for the effect of spin probe concentration, referred to as [trityl] by applying linear relation between [trityl] and relaxation rate (Fig. 3b). It should be noted that relaxation rates of spin probe in water show weaker and non-linear dependence on [trityl] (data not shown). The reasons for that are under investigation. There was no significant difference between the slopes of the linear dependences of R_1e and R_2e on pO₂. However, there was a dramatic difference in the slopes of the dependences of R_1e and R_2e on spin probe concentration, referred to as [trityl] (Fig. 3b). R_2e vs [trityl] had nearly five times greater slope than did R_1e.

Figure 3.

Relaxation rates of OX063 dissolved in saline at 37°C.

a. [trityl] corrected dependence (concentration independent) of relaxation rates on pO₂. The data were obtained using 0.46 mM sample. The relaxation rates were then extrapolated to zero [trityl] by subtracting 0.46mM·0.165·10⁶ s⁻¹/mM =76·10³ s⁻¹ for R_2e, and 0.46mM·36.3·10³ s⁻¹/mM =17·10³ s⁻¹ for R_1e. Best fit: R_1e = 8.9·10³ s⁻¹/torr * pO₂ +1.6·10⁵ s⁻¹; R_2e = 8.9·10³ s⁻¹/torr * pO₂ + 1.6·10⁵ s⁻¹.

b. [trityl] dependences of relaxation rates R_1e and R_2e. Best fit: R_1e = 36.3·10³ s⁻¹/mM * [trityl] +0.16·10⁵ s⁻¹; R_2e = 0.165·10⁶ s⁻¹/mM [trityl] + 0.16·10⁵ s⁻¹.

The 95% confidence intervals for fit parameters are: ±1·10⁴ s⁻¹ for offsets, ±1.2·10² s⁻¹/torr for O₂ and ±7·10³ s⁻¹/mM for concentration proportionality coefficients.

The quantitative dependence of R_2e on [trityl] is much weaker than on pO₂ or O₂ concentration, referred to as [O₂]. At 37 °C, [O₂] in saline solution is 219 µM at 159 torr or ~1.4 µM/torr. Thus, the sensitivity of trityl R_2e to [O₂] is ~6.3·10⁶ s⁻¹/mM, but only 0.165·10⁶ s⁻¹/mM for [trityl].

The results of R_1e imaging and relaxation measurements of phantoms with 0% and 9.3% O₂ (71 torr pO₂ at 37 °C and atmospheric pressure) are summarized in Table 3. These [O₂] bracket relevant values from pO₂ studies of hypoxia in animals (15,34). The relaxation times determined under non-imaging conditions are given in the footnotes of the table. No significant difference between average relaxation times determined under imaging and non-imaging conditions was found.

Table 3.

Precision of images estimated as a standard deviation of relaxation times in an image of homogeneous phantom. Images are obtained in 10 minutes on phantom containing 1 mM OX063 dissolved in normal saline at 37°C. No [trityl] correction is applied.

	0% pO2			9.3% pO2
Pulse Sequence	Average T2e or T1e [µs]	Standard deviation of T2e or T1e [µs]	Standard deviation of R2e or R1e [x 103 s−1]	Average T2e or T1e [µs]	Standard deviation of T2e or T1e [µs]	Standard deviation of R2e or R1e [×103 s−1]
TR measurements (T2e and R2e)
2pESE	3.1	0.2	7.9*	1.25	0.07	46*
SLR measurements (T1e and R1e)
IRESE	5.0	0.3	7.9*	1.31	0.15	90*

Non-imaging relaxation times: for 0% O₂ sample T_2e = 2.99 µs (two-pulse ESE), T_1e = 4.82 µs (IRESE); for 9.3% O₂ sample T_2e = 1.24 µs (two-pulse ESE), T_1e = 1.33 µs (IRESE).

^*7.9·10³ s⁻¹ corresponds to 0.9 torr; 46·10³ s⁻¹ corresponds to 5 torr; 90·10³ s⁻¹ corresponds to 10 torr.

Residual dependence of R_1e on [trityl] is likely due to spectral diffusion, the interaction of excited spins with the other spins whose EPR frequencies are beyond the excitation bandwidth (20). Such spins may be from trityl molecules containing ¹³C (1.109% natural abundance) producing hyperfine lines split up to 12 G from the central line, (35) considerably beyond the excitation bandwidth of the pulse sequence. The results of R_2e measurements using ESE are given in the table for comparison. To estimate errors, the standard deviation of the relaxation times was estimated for homogeneous phantom voxels, excluding regions with edge artifacts. The R_2e image had a standard deviation smaller than that of the R_1e imaging method. However, for 0% O₂ the standard deviation of the R_1e image approached that of the R_2e image.

R_1e in the tumor of a live animal shows far more freedom from [trityl] relaxation

For a demonstration of the robustness of R_1e imaging on a live animal tumor (Fig. 4b), we compared the IRESE image slice with the same slice from the R_2e image (Fig. 4a), obtained from a two-pulse ESE image on the same animal 10 minutes later. The brighter, richly colored areas in both of the images are well oxygenated, whereas the darker, more intensely blue areas are hypoxic. The tumor outline from a registered T₂-weighted MRI is shown in both the images as red contours. The outlines and general oxygenation patterns are very similar. However, absolute voxel pO₂ values differ considerably. Figure 4c presents the [trityl] image obtained with two-pulse ESE. Figures 4d and 4e are histograms of the R_2e and R_1e relaxation rates, respectively. The red-colored histograms are from the voxels enclosed by the red tumor contours in Figs. 4a–c. The blue histogram bars are from the leg tissue outside the tumor contours. They are added to the tumor histogram bars so that the ultimate height of the blue plus red histogram bars represents the total number of image voxels with the indicated relaxation rate.

The mode of the overall R_1e distribution from the tumor shown in Fig. 4e is nearly 0.35·10⁶ s⁻¹ smaller than that from the R_2e image. In the distributions of the R_1e and R_2e, there are two components with different modes and widths – the tumor and the residual leg area. The sharper component with the lower mode, localized primarily in the tumor area, is more clearly distinguished in the R_1e image. That component is associated with the narrower distribution of relaxation rates expected in the hypoxic tumor area and shows the improved performance of R_1e imaging. The slowest R_1e observed in the animal experiments (Fig. 4e) are very close to the rates found in deoxygenated phantoms. This is expected because tumors are frequently hypoxic. R_2e rates, however, are ~0.35·10⁶ s⁻¹ higher than R_1e rates (Fig. 4f). Because the relaxation rates due to different mechanisms are additive, the shift of the distribution toward higher values indicates that R_2e images are more susceptible to O₂-independent relaxation than is R_1e. At room temperature and at a negligible [trityl] the R_2e and R_1e are similar.

Ultimately, the advantage of R_1e-based pO₂ imaging over R_2e-based pO₂ imaging lies in its reduced susceptibility to confounding variation from [trityl]-induced relaxation. We demonstrated this by artificially increasing [trityl] by increasing the rate of trityl infusion to an animal and assessing the changes in the relaxation rates in selected regions. Although the local tissue oxygenation in subvolumes of tumors is known to vary (36,37), we assumed the average tissue oxygenation during the experiment to be constant. We justify this assumption below based on the results. In the absence of changes in oxygenation all changes in the relaxation rates should be from [trityl] relaxation effect. As a surrogate for the local [trityl], we used the voxel signal intensity obtained from R_2e images normalized to the intensity of voxels in a 1 mM phantom. The voxel signal intensity is R_2e-independent because the spin echo amplitude was extrapolated to τ=0. In Figure 5, the change in relaxation rates was imaged by alternating ESE and IRESE images. The [trityl], estimated from the ESE images, in animal tissues was stabilized prior to imaging by comparison of consecutive image intensities. Two regions of the image, one directly in the tumor volume and the other just outside the tumor volume in a well-perfused region, were compared for each image type. The regions consisted of a cube of 3×3×3=27 voxels, as seen in the leftmost image which shows a sagittal slice including the regions. The R_2e and R_1e are plotted to the right of the image slice from each region. The [trityl] dependence coefficients of R_2e and R_1e are indicated along with their uncertainties determined from the scatter of values about the regression lines. The R_2e values, for each location, depend strongly on the apparent [trityl] while the R_1e values are independent of [trityl] to within the value uncertainties shown. The effect of [trityl] on R_2e is considerably stronger than the effect of cycling hypoxia (36,37) and, more importantly is observed not only in tumor but in all anatomic areas.

Discussion

This work demonstrates that R_1e images have nearly an order of magnitude reduced sensitivity to self-relaxation in vivo and, thus, higher accuracy. The precision of R_1e and R_2e images is very similar, especially at low [O₂]. Figure 5 shows that R_2e does indeed depend on [trityl] in vivo. The trend is clearly visible because the range of [trityl] in Figure 5b is far larger than typically found in in vivo images. The slopes of R_2e [trityl] dependences in Fig. 5 ranging from 0.2·10⁶ to 0.9·10⁶ s⁻¹/mM are considerably stronger than that found in the phantoms (0.165·10⁶ s⁻¹/mM). In part, this may be due to our underestimation of the real spin probe concentration. The tri-acid spin probe cannot penetrate the cell membrane at physiologic pH (16). Its extracellular distribution volume can be as small as 30% of the tissue volume (38) increasing the actual concentration by a factor of three relative to that based on voxel intensity. The [trityl] dependence of R_1e is five times smaller in phantoms and is undetectable in vivo, Fig. 5, areas [1] and [2]. The pO₂, inferred from the R_1e images in each voxel are, to within the uncertainty, accurate and are the absolute measurements of tissue or tumor oxygenation.

As noted above, virtually all in vivo EPR O₂ images have exploited sensitivity of R_2e to local pO₂, except for one qualitative study (18). The present study demonstrates, for the first time, pulse EPR O₂ imaging in live animals that is sensitive to the spin probe electron R_1e. The long relaxation times of trityls and the O₂-induced changes in R_1e are much more favorable than with the nitroxides used in the past as an infusible spin probe for O₂ imaging in vivo (39,40). R_1e-based O₂ imaging is superior to R_2e or R*_2e imaging modalities because of the reduced sensitivity to [trityl]. The [trityl] in vivo is difficult to control and, given the unknown volume in each voxel occupied by the extracellular trityl spin probe, is difficult to determine accurately. The accuracy of R_1e O₂ images is thus improved by virtually eliminating sensitivity to trityl concentration.

Conclusions

R_1e EPR O₂ imaging with trityl spin probes is feasible and has the same precision as R_2e imaging. The absolute accuracy of this method is superior to R_2e imaging because of the smaller confounding effect of environmental parameters other than O₂ on R_1e. Therefore, R_1e EPR imaging should be considered as a primary method for oxymetry. Ten-minute images of a 1 mM hypoxic sample provide 1 torr pO₂ resolution with 1–1.5 mm spatial resolution (14,19).

The improved accuracy of the O₂ images, presented here, should further enhance animal studies of tissue and tumor oxygenation. The low radiofrequency at which these studies have been undertaken and the large animal tumors already studied with R_2e pO₂ images (29) argue for the distinct possibility for measurements in human tumors. pO₂ images can direct therapy of cancers with radiation. The enhanced accuracy of R_1e pO₂ imaging shown here argues strongly for this technique to be incorporated in future animal and perhaps human images.