Tumor oxygenation predicts cancer therapy response and malignant phenotype. This has spawned a number of oxymetries. Comparison of different oxymetries is crucial for the validation and understanding of these techniques. Electron paramagnetic resonance (EPR) imaging is a novel technique for providing quantitative high-resolution images of tumor and tissue oxygenation. This work compares sequences of tumor pO2 values from EPR oxygen images with sequences of oxygen measurements made along a track with an Oxylite oxygen probe. Four-dimensional (three spatial and one spectral) EPR oxygen images used spectroscopic imaging techniques to measure the width of a spectral line in each image voxel from a trityl spin probe (OX063, Amersham Health R&D) in the tissues and tumor of mice after spin probe injection. A simple calibration allows direct, quantitative translation of each line width to an oxygen concentration. These four-dimensional EPR images, obtained in 45 minutes from FSa fibrosarcomas grown in the legs of C3H mice, have a spatial resolution of ∼1 mm and oxygen resolution of ∼3 Torr. The position of the Oxylite track was measured within a 2-mm accuracy using a custom stereotactic positioning device. A total of nine images that involve 17 tracks were obtained. Of these, most showed good correlation between the Oxylite measured pO2 and a track located in the tumor within the uncertainties of the Oxylite localizability. The correlation was good both in terms of spatial distribution pattern and pO2 magnitude. The strong correlation of the two modalities corroborates EPR imaging as a useful tool for the study of tumor oxygenation.

The importance of oxygen and its absence (hypoxia) in cancer biology has grown in recent years. Hypoxia has classically been known to protect cells from radiation damage and is a source of tumor resistance to radiation (13). Recently, this role has been augmented by its role as a promoter of mutagenesis, proneoplastic transformation (4), and angiogenic signaling (5). Hypoxia seems to be a common characteristic of solid tumors, particularly large tumors (6). Although the transformation from normal to malignant tissue has been shown to proceed through multiple genetic alterations, in a stepwise fashion, it is common, in the clinic, to observe a substantial change in the biology of a solid tumor with a sudden acceleration of local growth and metastasis. This change may be due to both genetic and epigenetic effects, likely through their interaction. Hypoxia may be one of the major epigenetic participants in this change. Direct measurements of pO2 in tumors show that low pO2 predicts increased risk of failure in patients with malignancies of the head and neck, uterine cervix, and connective tissues treated with radiation or surgery (79). Hypoxia seems to predict malignant phenotype in patients whose cancers have been treated surgically (8). For example, in patients with uterine cervix carcinoma (8) and sarcomas (7), hypoxia in the primary tumor was correlated with subsequent development of distant metastases. Moreover, hypoxia is associated with both variable and significant response to chemotherapy, for example (10).

With the advent of modern radiation therapy delivery, intensity-modulated radiation therapy (11), it has become possible to subtly sculpt radiation dose distributions to avoid radiation sensitive anatomic structures. However, the region ascribed to the tumor-bearing tissue is treated homogeneously. This reflects our ignorance about regional variation in sensitivity within the tumor. If areas with high fractions of hypoxia could be defined, current technology could deliver substantially higher doses of radiation to those more hypoxic and resistant portions. (12) Present anatomic images of high resolution do not show changes in tissue physiology. A new, high-resolution, quantitative oxygen image is necessary to define the hypoxic regions to which these higher doses are delivered within the target volume. Such an oxygen image may provide at least an alternative and, because it is quantitative, a more repeatable and higher-resolution oxygen study than positron emission tomography scans to show response to therapy.

Because of its clinical importance, several techniques for measurements of oxygenation status in tissues and tumors of living animals and humans have been developed. Radiobiological hypoxic fractions (i.e., the proportion of clonogenic cells in a tumor, which are resistant to radiation due to hypoxia; ref. 13) provide a measure of tumor oxygenation. Selective retention of compounds that are reduced in hypoxic environments correlates with hypoxic fractions (14, 15) and can generate images of tumor hypoxia in living human subjects (16, 17). Oxygen concentrations in human tumors were measured with polarographic electrodes (8, 9). Point measurements of oxygen concentrations have been made with fluorescence quenching (18, 19) and particulate electron paramagnetic resonance (EPR) spectroscopy (20). Oxygen images have been obtained with fluorescence quenching (21), Overhauser enhanced magnetic resonance imaging (22, 23), 19F magnetic resonance imaging (24), and EPR imaging (25, 26). Noninvasive, three-dimensional tumor oxygen maps would improve clinical radiation treatment by enabling modulation of the radiation dose according to spatial information on tissue oxygenation.

Because each technique for hypoxia measurement emphasizes particular aspects of the cancer microenvironment, comparison of these techniques is crucial. Comparison both validates a new technique and helps understand more precisely what aspect of oxygenation physiology each is measuring. Hypoxic fractions have variably agreed (27) and disagreed (28) with oxygen electrode measurements. Disagreement may be due to the needle electrode insensitivity to necrosis. Comparisons of pooled measurements of oxygen electrodes and fiber optic fluorescence quenching have shown overall good cross validation (18, 19). Comparison of data from different imaging modalities or point measurements and images has been more difficult to obtain because of the difficulty in both spatial and temporal coregistration.

EPR oxygen imaging is a powerful new imaging technique that provides 3 Torr or better resolution in pO2 and better than 1-mm spatial resolution (26). The work presented here represents an effort to understand the relationship between pO2 images obtained using EPR spectroscopic imaging and Oxylite fiber optic point measurement of fluorescence quenching by oxygen. The difficulty in localizing the measurement in the EPR oxygen image was overcome by use of a stereotactic holder for the Oxylite fiber optic probe. Oxylite measurements were obtained in the EPR imager, just after an EPR tumor oxygen image with as little perturbation of the tumor as possible. The good agreement between Oxylite and the EPR oxygen image provides a strong cross-validation.

Spin probe. The spin probe used for the EPR imaging was 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, molecular weight = 1,427), a kind gift from Nycomed Innovations (Malmo, Sweden). It has a single inhomogeneously broadened spectral line with a peak line width of 16.5 μT in hypoxic solutions. The presence of oxygen further broadens the spin packet line by 0.054 μT/Torr. Effects of other free paramagnetic species, self-broadening, viscosity broadening, and power saturation are accounted for via an oxygen calibration (26). OX063 was dissolved in distilled water at a concentration of 0.25 g/mL and cleared of any undissolved particles using a 40-μm filter; 0.2 mL of this solution (1.4 mmol/kg body weight) was injected i.v. for a mouse weighing 25 g, which corresponds to one fifth of the LD50 for the spin probe (23). This triacid salt solution was slightly hypertonic and was well tolerated. The spin probe was retained in the tumor with much longer half-life (∼30 minutes) than that in other tissues.

Animal preparation. A total of nine tumors were used for the comparison of EPR imaging with Oxylite measurement. FSa cells (generously provided by Kathryn Mason, M.D. Anderson Hospital, Houston, TX) were injected i.m. in the right hind legs of 6-to 8-week-old C3H/HeN:Hsd (Harlan Sprague-Dawley, Indianapolis, IN) female mice so that the tumors grew to be ∼10 mm in diameter in 10 days. Animals were both immobilized and anesthetized for both Oxylite and EPR oxygen image measurements. This minimized animal movement during the procedures. Anesthesia involved i.p. ketamine (130 mg/kg) and xylazine (6 mg/kg) in 21.9 ± 0.8 g mice with upkeep of 20% this dose every half hour. In the EPR imager, the animal was immobilized by waterproof tape (WET-PRUF, Kendall, Mansfield, MA), and the tumor-bearing leg was placed in the 16-mm-diameter, 15-mm-thick loop of a loop-gap resonator. Tumor spatial stability and reproducibility were maintained by injecting hydrophilic vinyl polysiloxane dental impression material (GC Dental Products, Kasugai, Japan) under the leg to form a cast. The skin temperature, measured with a Physitemp (Clifton, NJ) digital needle probe thermometer, was kept at 33.5°C using opposed heating lamps. The spin probe rapidly accumulates in the mouse bladder (beginning within a minute after injection) and creates a source of confounding signal and artifact. To eliminate this, we flushed the bladder at 15 mL/h using a double-lumen urethral catheter and Harvard 22 syringe pump (Harvard Apparatus, Holliston, MA; ref. 29).

After placing the animal in the EPR imager and immobilizing the leg in the resonator, the EPR image was taken first. Immediately afterwards, without removing the animal, an Oxylite measurement was done. The interval between the beginning of imaging and the end of the Oxylite measurement was ∼2 hours. The complete experiment, including the animal preparation (∼45 minutes), could be up to 3 hours long. We carefully maintained a steady level of anesthesia and monitored the breathing rate and temperature during the experiment to ensure that the oxygenation level of the tissues was similar for EPR imaging and Oxylite measurements.

All animal experiments were done according to the USPHS “Policy on Humane Care and Use of Laboratory Animals,” and the protocols were approved by the University of Chicago Institutional Animal Care and Use Committee (ACUP No. 69681). The University of Chicago Animal Resources Center is an Association for Assessment and Accreditation of Laboratory Animal Care–approved animal care facility.

EPR imaging. Four-dimensional EPR images were obtained using a spectroscopic imager with an air-core magnet and Anderson type gradient coils (30) operating at 250 MHz and 9 mT as described previously (31). The Zeeman field was modulated at 4.98 KHz with overmodulation. We used an accurate line shape simulation that allows operation with high modulation amplitude (17 μT) to increase signal-to-noise ratio in images without sacrifice in the determination of the intrinsic line width (32, 33). The quality factor (Q) of the resonator was ∼280 without the tumor and 200 with the tumor installed. The power delivered to the resonator was 25 μW and well below the saturation level (200 μW).

EPR oxygen images were acquired and reconstructed using a four-dimensional tomographic spectroscopic imaging technique. This involves the acquisition of spectra with fixed gradients, referred to as projections. Each projection involved the acquisition of 256 field points scaled to uniformly populate the field interval determined by the gradient as has been described previously (34). The projections were subjected to a Gaussian filter whose width was four points and then subsampled to 64 points. A field of view of 3 cm was used. This determined our voxel size to be (3√2 cm/64) or 0.66 mm in the spatial direction. This voxel size determination has been discussed extensively elsewhere (26, 34).

Projections were obtained with 16 gradient amplitudes and 8 evenly spaced polar and azimuthal direction angles. An image required ∼45 minutes. The low signal-to-noise ratio at higher gradient projections were partially compensated for by signal averaging of multiple field sweeps, up to 10 averages for the highest gradient. Each sweep was obtained with 256 field points with acquisition time of 3 ms/point, 3 ms time constant, and 12-dB/octave filter using a SR830 lock-in amplifier (Stanford Research Systems, Sunnyvale, CA). The spatial field of view and the spectral field of view were 3 cm and 0.1 mT, respectively.

Spectroscopic image reconstruction used filtered back projection as described previously (26, 34, 35). All image voxels with spin probe amplitude ≤15% of the maximum spectral amplitude in the image were discarded. This 15% threshold was used to avoid fitting voxels with no signal or very poor signal. Small areas within several images had maximum intensities that did not survive this threshold. The oxygen values for these areas were set to zero. It should be noted that the Oxylite tracks were aimed toward the periphery of the tumor where such zero or very low signal regions were minimal.

The oxygen maps reconstructed and fitted from the images have a pO2 resolution of ∼3 Torr and a measured spatial resolution of ∼1 mm. Oxygen partial pressure resolution is limited by the accuracy of the spin probe spectral line width measurement. The actual source of the uncertainty is a complicated function of signal noise, sampling frequency, and the various forms of data filtering used in the reconstruction. In homogenous phantom imaging experiments, we have found that line width scatter is 0.16 μT, which is equivalent to 3 Torr via the oxygen versus line width calibration, which is discussed in ref. (26). Each image has a voxel size of 0.66 mm3. Thus, not only are the pixels shown in an image slice (0.66 mm)2, but also the slice is 0.66 mm thick.

Oxylite measurement. We used an Oxylite 2000 (Oxford Optronix, Oxford, United Kingdom) consisting of a 230-μm-diameter optical fiber for oxygen measurement and a thermocouple with fine wire leads. The oxygen probe tip (250-300 μm in diameter) has Ruthenium-III-(Tris)-chloride embedded in silicone polymer. It has a blue light–emitting diode that generates light pulses to induce fluorescence from the ruthenium luminophor. The oximetry is based on the principle that the lifetime of the fluorescent pulse is inversely proportional to the oxygen tension in the tip. A temperature sensor (T-type thermocouple with polyurethane coating) is attached in close proximity to the oxygen probe tip to correct for the temperature dependence of the fluorescence quenching. As delivered, the optical fiber can be moved into its Luer holder by as much as 4 mm. To fix the probe position more substantially to the holder, we applied the polysiloxane dental material between the probe and Luer fitting, which does not affect the calibration of the Oxylite (data not shown).4

4

Oxford Optronix, personal communication.

Immediately after the EPR imaging while the animal was still in the EPR imager and under anesthesia, the sterotactic frame for location of the Oxylite probe tip was rapidly and easily installed. The skin of the tumor was punctured with a 21-gauge needle (0.8 mm diameter) mounted on the frame, which, in turn, guided the Oxylite probe tip. Once the skin was punctured, the fiber optic probe was stiff enough to proceed into the tissue in the direction defined by the stereotactic positioning device (Fig. 1). The Oxylite probe was inserted into the puncture wound and then fully through the tumor to ∼1 cm depth. Occasionally, low pO2 readings were obtained at the surface opposite the introduction of the Oxylite due to trauma with the initial introduction of the probe (36). Several minutes were initially allowed to get a steady reading (19, 37). pO2 was measured after each 1-mm retraction of the probe, eventually to the initial insertion point (Fig. 1). Flexion in the probe gave an uncertainty in the lateral position of the Oxylite track to within 2 mm (see below). Variation in the location of the resonator in the direction of its axis (y direction in Fig. 1) in the gradient coordinate system gave a longitudinal uncertainty of 2 mm as well.

Fig. 1.

Diagram and photograph of the Oxylite probe, mounted in the stereotactic frame and penetrating the FSa tumor in the hind leg of a C3H mouse immobilized in the EPR resonator. The Oxylite probe was advanced to the length of the tumor. Spatial coordinates of probe tip can be controlled within 2-mm accuracy. For scale, the diameter of the tumor bearing inductive element of the resonator (the hole through which the tumor bearing leg extends) is 16 mm. A, diagram of the stereotactic mount for the delivery of the Oxylite probe. B, photograph of the same showing the Oxylite probe penetrating the tumor.

Fig. 1.

Diagram and photograph of the Oxylite probe, mounted in the stereotactic frame and penetrating the FSa tumor in the hind leg of a C3H mouse immobilized in the EPR resonator. The Oxylite probe was advanced to the length of the tumor. Spatial coordinates of probe tip can be controlled within 2-mm accuracy. For scale, the diameter of the tumor bearing inductive element of the resonator (the hole through which the tumor bearing leg extends) is 16 mm. A, diagram of the stereotactic mount for the delivery of the Oxylite probe. B, photograph of the same showing the Oxylite probe penetrating the tumor.

Close modal

The probes have a finite practical usage time, which is limited by photo bleaching to ∼24 hours of use. At each experiment, we calibrated the probe by bubbling pure Argon and three gas mixes of N2 and O2 (3.0%, 5.6%, and 9.3%) through water at 37°C. The atmospheric pressure logged at an internet weather service (Weather Underground, http://wunderground.com), corrected for our altitude (183 m), was used for the calibration. Most of the tumors underwent two Oxylite track measurements that took up to 1 hour. We obtained a total of 17 track measurements from nine tumors.

Selecting the location of the Oxylite in the oxygen image. It was assumed that the Oxylite probe tip moved through the tumor parallel to the resonator axis, the y axis in Fig. 1A and subsequent figures. Although the position of the sterotactic frame can be controlled within 1-mm accuracy, the track of the probe tip in the tissue may deviate from the straight line parallel to the resonator axis. The advancing and retraction of the fiber optic probe can distort the longitudinal position of the tissue by compressing it or stretching it by ∼1 mm. It is difficult to localize the probe tip due to bending and curvature of the fiber optic probe. This curvature is shown in Fig. 2, a high-resolution X-ray image of the mouse leg tumor immobilized as in the Oxylite-EPR image comparison. The overall uncertainty in the localization of the Oxylite track is ∼2 mm. This uncertainty was accounted for by searching the adjacent track space in the EPR oxygen image within this limit to find the track with the best match to the Oxylite pO2 profile. Given the high gradients in the oxygen distribution and the high spatial resolution of the EPR oxygen images, it was very difficult to locate the Oxylite probe tip to within the image resolution. We found that within the cylinder of uncertainty of the Oxylite probe tip location in the EPR image, we could reliably identify a sequence or “track” of EPR oxygen image values that agreed with the Oxylite values. Although this may seem circular, we felt that it was reasonable to use the track whose oxygen values best agreed with the Oxylite. We tested this by trying to find matches to an Oxylite track in distant portions of the tumor and failed to find such matches.

Fig. 2.

High-resolution X-ray image of C3H mouse with Oxylite probe in the FSa tumor. It was taken with an ultra high resolution Faxitron MX-20, X-ray device with maximum X-ray energy of 28 kV, 1 R exposure, and a focal spot size of 20 μm. The position of the needle is controlled to within a millimeter by the stereotactic positioning device. The needle guides the Oxylite to the tumor and its sharp tip breaks the skin. The Oxylite protrudes from the needle and is seen curving away from the direction of the needle. The actual optic fiber is seen as a faint shadow accompanying a thin, well-resolved filament. The filament is the lead for the thermocouple, which provides the temperature correction for the Oxylite measurement.

Fig. 2.

High-resolution X-ray image of C3H mouse with Oxylite probe in the FSa tumor. It was taken with an ultra high resolution Faxitron MX-20, X-ray device with maximum X-ray energy of 28 kV, 1 R exposure, and a focal spot size of 20 μm. The position of the needle is controlled to within a millimeter by the stereotactic positioning device. The needle guides the Oxylite to the tumor and its sharp tip breaks the skin. The Oxylite protrudes from the needle and is seen curving away from the direction of the needle. The actual optic fiber is seen as a faint shadow accompanying a thin, well-resolved filament. The filament is the lead for the thermocouple, which provides the temperature correction for the Oxylite measurement.

Close modal

Statistical analysis. There are a number of ways to analyze the correspondence between the Oxylite point oxygen measurements and the corresponding voxel oxygen measurements from the EPR oxygen image. We believe that the most powerful means by which to view the association between the two methods is what we have presented, plotting the oxygen values as a function of depth in the tumor measured with each modality. This highlights the spatial correspondence between the measurements. Statistical measures can obscure such spatially distinct correlations. However, the magnitude of the correlation between the EPR oxygen image values and the corresponding Oxylite oxygen values can be estimated using the Pearson product moment correlation coefficient R (38). Note that R values were obtained from each of the 17 sets of paired oxygen track measurements. The correlation was also obtained for the whole set of 17 comparison measurements. The overall distributions of oxygen measurements are compared in separate, adjacent graphs. One aspect of our data is the large number of zero values obtained from EPR oxygen image. This leads, in certain cases, to a rather non-normal distribution of values for which the correlation coefficient may not be the best measure. In addition, as discussed in Bland and Altman (39), when assessing agreement between two methods of measurement, it may be preferable to simply focus on the magnitude of the differences between them. Therefore, as an additional measure of agreement between the Oxylite and the EPR oxygen image measurement of oxygen partial pressure, we consider the point by point absolute difference between them. The mean and median values of these absolute differences were calculated for each track.

Additional data. Because of concern about the repeatability of these measurements, a second set of data was obtained. Ten tumors were imaged. In each tumor, two Oxylite tracks, for a total of 20 tracks, were obtained in a manner identical to that described above. These tracks were cast into the EPR oxygen image using the same methods described above, and the same analysis was done. The statistical characteristics of the two groups of measurements were compared by applying the Student's t test to the distributions of both Pearson product-moment correlation coefficients and mean absolute differences per track.

The experimental set up is diagramed in Fig. 1A. A C3H leg bearing an FSa fibrosarcoma is shown positioned in the EPR resonator. Figure 1A also shows the mouse leg and the delivery end of the stereotactic needle positioning frame and the coordinate system defined by the EPR imager in the top left corner. The Oxylite fiber optic probe is shown piercing the tumor. Note that Oxylite measurements were obtained without moving the mouse leg after the EPR imaging. Figure 1B shows a photograph of the set up.

The oxygen image of the anatomy inside the resonator, including the tumor, was obtained immediately before the Oxylite measurement and is shown in Fig. 3. The oxygen image in Fig. 3 shows two slices parallel to the axis of the resonator. Their intersection is a line roughly parallel to the tibia. These are quantitative images and the color of the image corresponds directly to a voxel pO2 measured in Torr or mm Hg. The very bright, rich-colored regions in Fig. 3 represent highly oxygenated regions (bottom center in the top left slice) or superficial skin (about the surface of the image) with pO2 approaching 75 Torr. The darker blue-colored regions of the image show hypoxic tumor.

Fig. 3.

Surface rendering of a tumor oxygen image from September 21, 2004, embedded in which are two orthogonal slices (0.66 mm thick) showing the quantitative oxygen values in the voxels contained in the slices. The surface rendering was derived from the EPR spectral amplitude image with a threshold at 15% maximum. The z-y slice is horizontal, roughly in the mouse coronal plane. The x-y slice is a vertical slice, roughly in the mouse sagittal plane. The intersection of the two planes is roughly parallel to the tibia and the axis of the resonator in Fig. 1. Darker blue, hypoxic areas; reds and yellows, higher oxygen values. The quantitative color bars are shown in Fig. 4.

Fig. 3.

Surface rendering of a tumor oxygen image from September 21, 2004, embedded in which are two orthogonal slices (0.66 mm thick) showing the quantitative oxygen values in the voxels contained in the slices. The surface rendering was derived from the EPR spectral amplitude image with a threshold at 15% maximum. The z-y slice is horizontal, roughly in the mouse coronal plane. The x-y slice is a vertical slice, roughly in the mouse sagittal plane. The intersection of the two planes is roughly parallel to the tibia and the axis of the resonator in Fig. 1. Darker blue, hypoxic areas; reds and yellows, higher oxygen values. The quantitative color bars are shown in Fig. 4.

Close modal

The coordinates of the Oxylite track in the image were determined by relating them to the measurement of the edge of the resonator with the Oxylite probe using the stereotactic positioning frame. This, in turn, assumed consistent and exact placement of the resonator in the image coordinate frame, which is determined by the gradients and the location of the gradient-producing coils. Photographs taken during the experiment provided a general region in the image coordinate system in which the “true” track coordinates lay. Based on the coordinates read from the positioning device, the estimate of the Oxylite probe track position was refined and cast into the EPR oxygen image as shown in Fig. 4. These tracks in the image were assumed to lie parallel to the axis of the resonator (the “Y” direction in Figs. 1, 3, and 4). This restrictive assumption simplified the analysis of the data but allowed the accumulation of positional discrepancy between the location of the Oxylite tip and the corresponding image voxel. On the other hand, more flexible assumptions about the needle path, compromised and complicated the analysis as to reduce its significance. An adjacent track is defined as any one of the eight lines in the image pointing in the “Y” direction next to the primary track.

Fig. 4.

Coordinates quantified refer to the image coordinates with zeros approximately in the center of the resonator and the tumor. A, tumor oxygenation sagittal slice with quantitative color maps taken from the four-dimensional EPR oxygen image of the tumor and leg shown in Fig. 3. Two adjacent tracks in the image (black and white), which correspond to the Oxylite track 2 from September 21, 2004. The color map provides a quantitative measure of the oxygen in each of the 0.66-mm3 voxels. B, plot of the oxygen values from the (•) Oxylite track and (○) the two adjacent black and white tracks from (A). The oxygen image “track” with the higher values corresponds to the black track in the image. The horizontal error bar placed at the bottom shows the uncertainty in needle localization (2 mm). C, the coronal slice shown in Fig. 3 with a quantitative color map. This is the perpendicular slice through the black track in (A).

Fig. 4.

Coordinates quantified refer to the image coordinates with zeros approximately in the center of the resonator and the tumor. A, tumor oxygenation sagittal slice with quantitative color maps taken from the four-dimensional EPR oxygen image of the tumor and leg shown in Fig. 3. Two adjacent tracks in the image (black and white), which correspond to the Oxylite track 2 from September 21, 2004. The color map provides a quantitative measure of the oxygen in each of the 0.66-mm3 voxels. B, plot of the oxygen values from the (•) Oxylite track and (○) the two adjacent black and white tracks from (A). The oxygen image “track” with the higher values corresponds to the black track in the image. The horizontal error bar placed at the bottom shows the uncertainty in needle localization (2 mm). C, the coronal slice shown in Fig. 3 with a quantitative color map. This is the perpendicular slice through the black track in (A).

Close modal

Figure 4A and C show two perpendicular planes in the oxygen image passing through the sequence of oxygen voxels shown with a black line. Examples of the pO2 readings both from Oxylite measurement and EPR image (the same slices shown in Fig. 3) are plotted in Fig. 4B. In this particular case, readings from adjacent tracks in the EPR oxygen image bracket the values from the Oxylite track. Referring back to Fig. 4A, one can see a region of high oxygen tension at the right end of the track. Moving to the left (toward negative Y location values), the track passes through very low pO2 values. It then passes just by the tip of a high oxygen protuberance, a region of substantial oxygen gradient. The high gradients seen in the oxygen image are reflected in the large variations in the Oxylite readings over 2 mm. The horizontal error bar placed at the bottom of Fig. 4B emphasizes the uncertainty in needle localization and the difficulty in extracting the point by point agreement between the two modalities.

Of the 17 tracks obtained from nine images, most showed good correlation between the Oxylite measured pO2 and the pO2 values from a track in the EPR oxygen image within the uncertainties of the Oxylite location. In most of the tracks where the product moment correlation R was not very high, one can still visualize the good spatial resemblance between the Oxylite and the oxygen image measurements. The Oxylite oxygen values with the oxygen values from closest EPR oxygen image track values are plotted in Fig. 5. The corresponding Pearson product moment correlation coefficients are displayed in Table 1. The variations in the Oxylite are reflected, for the most part, in the variations of the “tracks” from the EPR oxygen image. Moreover, as seen in Table 1 several of the tracks with low R values had very small average and median absolute differences. For example, the second to lowest mean absolute difference and the lowest median absolute difference (0.2 Torr) occurs in track 1 from September 21. This indicates very good agreement in values. However, the R value is only 0.28. A discrepancy in the opposite direction occurs between the two measures in the track from October 1 showing the largest mean and median absolute pO2 differences but an R value of 0.86.

Fig. 5.

Measurements of pO2 from both Oxylite and EPR imaging are plotted. Of the 17 tracks obtained from nine images, all showed relatively good correlation between the two modalities. The abscissa measures the distance from the entry point of the Oxylite into the tumor relative to the coordinate center of the image.

Fig. 5.

Measurements of pO2 from both Oxylite and EPR imaging are plotted. Of the 17 tracks obtained from nine images, all showed relatively good correlation between the two modalities. The abscissa measures the distance from the entry point of the Oxylite into the tumor relative to the coordinate center of the image.

Close modal
Table 1.

Statistical variables for the distribution of oxygen values shown in Fig. 6 measured using the Oxylite and the associated track from the EPR oxygen image

Median (mm Hg)Mode (mm Hg)Mean (mm Hg)
Oxylite 3.6 12.4 
EPRI 3.5 16.0 
    
Track
 
R
 
Median absolute difference (Torr)
 
Mean absolute difference
 
Track 2, Aug 24, 2004 0.38 9.2 13.2 
Track 1, September 17, 2004 0.94 0.3 4.3 
Track 2, September 17, 2004 0.78 9.9 11.4 
Track 1, September 21, 2004 0.28 0.2 2.4 
Track 2, September 21, 2004 0.81 3.6 7.9 
Track 1, September 30, 2004 0.70 4.5 10.6 
Track 2, September 30, 2004 0.46 6.6 17.9 
Track 1, October 1, 2004 0.86 20.8 34.3 
Track 2, October 1, 2004 0.87 3.3 5.5 
Track 1, October 5, 2004 0.53 9.7 17.0 
Track 2, October 5, 2004 0.97 1.1 1.8 
Track 1, October 7, 2004 0.85 2.0 6.9 
Track 2, October 7, 2004 0.14 3.3 13.5 
Track 1, October 8, 2004 0.21 1.8 11.0 
Track 2, October 8, 2004 0.29 3.0 7.3 
Track 1, October 11, 2004 0.95 3.8 4.2 
Track 2, October 11, 2004 0.85 3.9 7.8 
Mean 0.64 ± 0.07 5.1 ± 1.3 10.4 ± 1.9 
    
Track 1, mouse 2, 13 December, 2005 0.48 7.3 13.6 
Track 2, mouse 2, 13 December, 2005 0.86 9.7 15.3 
Track 1, mouse 1, 14 December, 2005 0.87 8.0 12.9 
Track 2, mouse 1, 14 December, 2005 0.71 15.3 18.3 
Track 1, mouse 2, 14 December, 2005 0.92 7.2 8.7 
Track 2, mouse 2, 14 December, 2005 0.66 8.6 15.8 
Track 1, mouse 1, 15 December, 2005 0.59 12.0 16.1 
Track 2, mouse 1, 15 December, 2005 0.55 9.5 22.6 
Track 1, mouse 2, 15 December, 2005 0.99 1.8 2.7 
Track 2, mouse 2, 15 December, 2005 0.56 9.8 18.4 
Track 1, mouse 1, 18 December, 2005 0.68 3.2 8.0 
Track 2, mouse 1, 18 December, 2005 0.88 4.9 10.5 
Track 1, mouse 2, 18 December, 2005 0.46 15.2 24.0 
Track 2, mouse 2, 18 December, 2005 0.56 3.7 14.4 
Track 1, mouse 1, 28 December, 2005 0.76 4.5 12.8 
Track 2, mouse 1, 28 December, 2005 0.68 12.9 13.3 
Track 1, mouse 1, 29 December, 2005 0.70 0.4 1.7 
Track 2, mouse 1, 29 December, 2005 0.34 0.4 2.7 
Track 1, mouse 2, 29 December, 2005 0.51 9.4 17.2 
Track 2, mouse 2, 29 December, 2005 0.60 12.6 19.0 
Mean 0.67 ± 0.04 9.1, ± 1.2 13.4 ± 1.4 
Median (mm Hg)Mode (mm Hg)Mean (mm Hg)
Oxylite 3.6 12.4 
EPRI 3.5 16.0 
    
Track
 
R
 
Median absolute difference (Torr)
 
Mean absolute difference
 
Track 2, Aug 24, 2004 0.38 9.2 13.2 
Track 1, September 17, 2004 0.94 0.3 4.3 
Track 2, September 17, 2004 0.78 9.9 11.4 
Track 1, September 21, 2004 0.28 0.2 2.4 
Track 2, September 21, 2004 0.81 3.6 7.9 
Track 1, September 30, 2004 0.70 4.5 10.6 
Track 2, September 30, 2004 0.46 6.6 17.9 
Track 1, October 1, 2004 0.86 20.8 34.3 
Track 2, October 1, 2004 0.87 3.3 5.5 
Track 1, October 5, 2004 0.53 9.7 17.0 
Track 2, October 5, 2004 0.97 1.1 1.8 
Track 1, October 7, 2004 0.85 2.0 6.9 
Track 2, October 7, 2004 0.14 3.3 13.5 
Track 1, October 8, 2004 0.21 1.8 11.0 
Track 2, October 8, 2004 0.29 3.0 7.3 
Track 1, October 11, 2004 0.95 3.8 4.2 
Track 2, October 11, 2004 0.85 3.9 7.8 
Mean 0.64 ± 0.07 5.1 ± 1.3 10.4 ± 1.9 
    
Track 1, mouse 2, 13 December, 2005 0.48 7.3 13.6 
Track 2, mouse 2, 13 December, 2005 0.86 9.7 15.3 
Track 1, mouse 1, 14 December, 2005 0.87 8.0 12.9 
Track 2, mouse 1, 14 December, 2005 0.71 15.3 18.3 
Track 1, mouse 2, 14 December, 2005 0.92 7.2 8.7 
Track 2, mouse 2, 14 December, 2005 0.66 8.6 15.8 
Track 1, mouse 1, 15 December, 2005 0.59 12.0 16.1 
Track 2, mouse 1, 15 December, 2005 0.55 9.5 22.6 
Track 1, mouse 2, 15 December, 2005 0.99 1.8 2.7 
Track 2, mouse 2, 15 December, 2005 0.56 9.8 18.4 
Track 1, mouse 1, 18 December, 2005 0.68 3.2 8.0 
Track 2, mouse 1, 18 December, 2005 0.88 4.9 10.5 
Track 1, mouse 2, 18 December, 2005 0.46 15.2 24.0 
Track 2, mouse 2, 18 December, 2005 0.56 3.7 14.4 
Track 1, mouse 1, 28 December, 2005 0.76 4.5 12.8 
Track 2, mouse 1, 28 December, 2005 0.68 12.9 13.3 
Track 1, mouse 1, 29 December, 2005 0.70 0.4 1.7 
Track 2, mouse 1, 29 December, 2005 0.34 0.4 2.7 
Track 1, mouse 2, 29 December, 2005 0.51 9.4 17.2 
Track 2, mouse 2, 29 December, 2005 0.60 12.6 19.0 
Mean 0.67 ± 0.04 9.1, ± 1.2 13.4 ± 1.4 

NOTE: In addition to Pearson product-moment correlation, coefficient R and mean and median absolute pO2 differences for all tracks presented in Fig. 5 and for those obtained in the additional data set are shown.

There are disagreements between the oxygen image values and those from the Oxylite. These are most often largest at the largest depth in the tumor, where the largest deviation of the actual Oxylite track from the assumed path direction is likely to be greatest. Given the rapid variation of the oxygen values in portions of these images, such positional deviations will translate into oxygen deviations. It is very difficult to overcome such deviations at the level of resolution of the image and the resolution of the oxygen gradients seen in the image.

Finally, in Fig. 6, we display the distribution of the oxygenation values from both the Oxylite and the oxygen image measurements. The mean, mode, and median values for these distributions are tabulated in Table 1. As can be seen, the robust statistical variables (mode and median) values are virtually identical, whereas means, difficult to apply to such a skewed distribution, are close.

Fig. 6.

Histogram of the oxygen concentrations shown in Fig. 5. A, histogram of oxygen concentrations from Oxylite measurements. B, histogram of oxygen concentrations from EPR imaging interpolated from the nearest voxels to the location of the Oxylite (i.e., the “track” reconstructed in the EPR oxygen image).

Fig. 6.

Histogram of the oxygen concentrations shown in Fig. 5. A, histogram of oxygen concentrations from Oxylite measurements. B, histogram of oxygen concentrations from EPR imaging interpolated from the nearest voxels to the location of the Oxylite (i.e., the “track” reconstructed in the EPR oxygen image).

Close modal

The correlation results and difference measures from the second set of data are also shown in Table 1. The mean of the Pearson product moment coefficients R, 0.64 from the first set and 0.67 from the second, are very similar and the difference is statistically insignificant (P = 0.71). The average of the mean absolute difference measures, 10.4 from the first set and 13.4 from the second set, are also not significantly different from one another (P = 0.2).

The pO2 distributions in Fig. 5 show good agreement between the Oxylite fiber optic probe pO2 and corresponding tracks localized in the EPR oxygen image both in terms of spatial distribution pattern and pO2 magnitude. The diameter of Oxylite probe tip (230 μm) is smaller than the 300-μm-diameter Eppendorf needle, but the method is essentially invasive. There can be traumatic change in local tissue oxygenation with the insertion of the probe.

Both EPR oxygen image slices in Figs. 3 and 4 and the data presented in Fig. 5 show the heterogeneity in the oxygenation of the mouse legs with implanted FSa tumors. The tracks passed through both tumor and muscle tissue. In 14 of 17 tracks shown in Fig. 5, extremely hypoxic areas of pO2 were found, often in the middle of the track. There are some well-oxygenated areas as well, with values of 40 Torr and sometimes more, most likely in the region of normal tissue.

Despite the overall good agreement, there are differences between the Oxylite tracks and the EPR oxygen image tracks. Note that seven of the images show small regions with large discrepancies between the Oxylite pO2 values and those from the EPR oxygen image. In all but one of these, the discrepancies occur at the most distal points in the Oxylite track. In five of the distal discrepancies, there are high EPR oxygen measurements, whereas the Oxylite measures low or zero oxygen values. There are at least two reasons why these differences might have occurred in this set of five tracks. The abscissa measures the distance from the entry point of the Oxylite into the tumor. (a) At the largest abscissa values, deepest in the tumor, the 200-μm glass fiber Oxylite probe is most likely to maximally deviate from the assumed straight line track. This could be as much as 2 mm as shown in Fig. 2, a very large distance in these oxygen images given the large oxygen gradients shown in Figs. 3 and 4. (b) Both of the September 30 images show very high oxygen concentrations at the surface of the tumor. These may be out of the range of reliability for the Oxylite.

Track 2 from September 17 shows a discrepancy between the Oxylite and the EPR oxygen image at the beginning of the track. Review of the EPR oxygen image shows very rapid variation of the oxygen values in tracks near the Oxylite. Slight misregistration can explain such discrepancies.

In track 1 from October 8, 2004, the trityl intensity was unusually low in the region of discrepancy. Thus, there is reason to suspect those data points in the EPR image in this very small region. This points out the richness of the information that is available from EPR images. There are other variables of the spectrum that measure the quality of the data. This allows assessment of situations where data are less than optimum or which subset of the data is not fully trustworthy.

However, the bulk of the data shown in Figs. 5 and 6 show a strong correlation between the Oxylite and the EPR oxygen image. The statistics used to describe the two different data sets show deviations and summary values that are less encouraging than one might wish from an ideal set of measurements, where the location of each measurement with each modality could be identified. This is beyond the scope of needle localization. It contaminates the statistics. However, statistics, even when applied pairwise to a single track, are only summaries of the data. Interestingly, we show cases where two reasonable statistical summaries disagree. The real data is what is shown in Fig. 5. There are regions of discrepancy seen there, but the patterns and the magnitude of the variation are highly consistent. This study supports EPR imaging as a useful tool for the study of tumor oxygenation in animal tumors. pO2 has been shown to strongly correlate with the radiation curability of tumors in humans (8). Nonetheless, there is good data showing disagreement between pO2 distribution measurements and classic hypoxic fraction measurement (28). A major strength of EPR imaging lies in the richness of the information available from it. In addition to providing noninvasive in vivo spatial oxymetry, the images provide maps of the intensity of the spin probe distribution. Analysis of spin probe concentrations and/or line width response to oxygen manipulation (e.g., via a carbogen challenge) may provide information allowing distinction between viable and necrotic regions. Future correlation with magnetic resonance imaging dynamic contrast enhancement and blood oxygen level-dependent contrast will further clarify this issue. The present 3-Torr oxygen resolution is less than ideal from the perspective of classic oxygen enhancement of radiation (40). However, most human measurements using Eppendorf electrode measurements show that a useful prognostic oxygenation value is 10 Torr. Under these circumstances, a 3-Torr resolution is quite useful. Moreover, new acquisition techniques and new spin probes should enhance this resolution (41). New narrower line spin probes will enhance spatial resolution in animals or allow scaling to human subjects at the present or close to the present spatial resolution. As the resolution of the technique improves, we expect much more insight into tumor physiology.

Grant support: NIH grants R01 CA 98575 and P41 EB 002034, Department of Defense grant DAMD17-02-1-0034, and North Atlantic Treaty Organization grant LST.CLG. 978628.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Note: M. Elas and K-H. Ahn contributed equally to this work.

We thank Drs. Ralph W. Weichselbaum and Charles A. Pelizzari for providing helpful comments and advice and Philip Schumm and Dr. Theodore Karrison for providing valuable advice concerning the statistical analysis of this data.

1
Schwarz G. Uber Desensibilisierung gegen Rontgen- und Radiumstrahlen.
Munchner Medizinische Wochenschrift
1909
;
56
:
1217
–8.
2
Holthusen H. Beitrage zur Biologie der Strahlenwirkung.
Pflugers Arch
1921
;
187
:
1
–24.
3
Thomlinson RH, Gray LH. The histological structure of some human lung cancers and the possible implications for radiotherapy.
Br J Radiol
1955
;
9
:
539
–63.
4
Graeber TG, Osmanian C, Jacks T, et al. Hypoxia-mediated selection of cells with diminished apoptotic potential in solid tumours.
Nature
1996
;
379
:
88
–91.
5
Risau W. Mechanisms of angiogenesis.
Nature
1997
;
386
:
671
.
6
Vaupel P, Fortmeyer HP, Runkel S, Kallinowski F. Blood flow, oxygen consumption, and tissue oxygenation of human breast cancer xenografts in nude rats.
Cancer Res
1987
;
47
:
3496
–503.
7
Brizel DM, Scully SP, Harrelson JM, et al. Tumor oxygenation predicts for the likelihood of distant metastases in human soft tissue sarcoma.
Cancer Res
1996
;
56
:
941
–3.
8
Hockel M, Schlenger K, Aral B, Mitze M, Schaffer U, Vaupel P. Association between tumor hypoxia and malignant progression in advanced cancer of the uterine cervix.
Cancer Res
1996
;
56
:
4509
–15.
9
Brizel DM, Dodge RK, Clough RW, Dewhirst MW. Oxygenation of head and neck cancer: changes during radiotherapy and impact on treatment outcome.
Radiother Oncol
1999
;
53
:
113
–7.
10
Skarsgard LD, Skwarchuk MW, Vinczan A, Kristl J, Chaplin DJ. The cytotoxicity of melphalan and its relationship to pH, hypoxia and drug uptake.
Anticancer Res
1995
;
15
:
219
–23.
11
Mundt A, Roeske J. Intensity modulated radiation therapy. Hamilton (Ontario): B.C. Decker; 2005. p. 544.
12
Chao KS, Bosch WR, Mutic S, et al. A novel approach to overcome hypoxic tumor resistance: Cu-ATSM-guided intensity-modulated radiation therapy.
Int J Radiat Oncol Biol Phys
2001
;
49
:
1171
–82.
13
Hall EJ. Radiobiology for the radiologist. Philadelphia: Lippincott Williams & Wilkins; 2000. p. 588.
14
Koch CJ, Stobbe CC, Baer KA. Metabolism induced binding of 14C-misonidazole to hypoxic cells: kinetic dependence on oxygen concentration and misonidazole concentration.
Int J Radiat Oncol Biol Phys
1984
;
10
:
1327
–31.
15
Raleigh JA, Chou SC, Arteel GE, Horsman MR. Comparisons among pimonidazole binding, oxygen electrode measurements, and radiation response in C3H mouse tumors.
Radiat Res
1999
;
151
:
580
–9.
16
Koh WJ, Rasey JS, Evans ML, et al. Imaging of hypoxia in human tumors with [F-18]fluoromisonidazole.
Int J Radiat Oncol Biol Phys
1992
;
22
:
199
–212.
17
Dehdashti F, Grigsby PW, Mintun MA, Lewis JS, Siegel BA, Welch MJ. Assessing tumor hypoxia in cervical cancer by positron emission tomography with 60Cu-ATSM: relationship to therapeutic response-a preliminary report.
Int J Radiat Oncol Biol Phys
2003
;
55
:
1233
–8.
18
Griffiths JR, Robinson SP. The OxyLite: a fibre-optic oxygen sensor.
Br J Radiol
1999
;
72
:
627
–30.
19
Seddon BM, Honess DJ, Vojnovic B, Tozer GM, Workman P. Measurement of tumor oxygenation: in vivo comparison of a luminescence fiber-optic sensor and a polarographic electrode in the p22 tumor.
Radiat Res
2001
;
155
:
837
–46.
20
Goda F, Liu KJ, Walczak T, Ohara JA, Jiang JJ, Swartz HM. In vivo oximetry using EPR and India ink.
Magn Reson Med
1995
;
33
:
237
–45.
21
Vanderkooi JM, Maniara G, Green TJ, Wilson DF. An optical method for measurement of dioxygen concentration based upon quenching of phosphorescence.
J Biol Chem
1987
;
262
:
5476
–82.
22
Golman K, Petersson JS, Ardenkjaer-Larsen JH, et al. Dynamic in vivo oxymetry using overhauser enhanced MR imaging.
J Magn Reson Imaging
2000
;
12
:
929
–38.
23
Krishna MC, English S, Yamada K, et al. Overhauser enhanced magnetic resonance imaging for tumor oximetry: coregistration of tumor anatomy and tissue oxygen concentration.
Proc Natl Acad Sci U S A
2002
;
99
:
2216
–21.
24
Mason RP, Shukla H, Antich PP. In vivo oxygen tension and temperature: simultaneous determination using 19F NMR spectroscopy of perfluorocarbon.
Magn Reson Med
1993
;
29
:
296
–302.
25
Kuppusamy P, Shankar RA, Zweier JL. In vivo measurement of arterial and venous oxygenation in the rat using 3D spectral-spatial electron paramagnetic resonance imaging.
Phys Med Biol
1998
;
43
:
1837
–44.
26
Elas M, Williams BB, Parasca A, et al. Quantitative tumor oxymetric images from 4D electron paramagnetic resonance imaging (EPRI): methodology and comparison with blood oxygen level-dependent (BOLD) MRI.
Magn Reson Med
2003
;
49
:
682
–91.
27
Horsman MR, Khalil AA, Nordsmark M, Grau C, Overgaard J. Measurement of pO2 in a murine tumour and its correlation with hypoxic fraction.
Adv Exp Med Biol
1994
;
345
:
493
–500.
28
Kavanagh MC, Sun A, Hu Q, Hill RP. Comparing techniques of measuring tumor hypoxia in different murine tumors: Eppendorf pO2 histograph, [3H]misonidazole binding and paired survival assay.
Radiat Res
1996
;
145
:
491
–500.
29
Haney CR, Ichikawa K, Parasca AD, et al. Reduction of image artifacts in mice by bladder flushing with a novel mouse double lumen uthrhreal catheter.
Mol Imaging
2006
;
5
:
1
–5.
30
Anderson W. Electrical current shims for correcting magnetic fields.
Rev Sci Instrum
1961
;
32
:
241
–50.
31
Halpern HJ, Bowman MK, Spencer DP, et al. An imaging radiofrequency electron spin resonance spectrometer with high resolution and sensitivity for in vivo measurements.
Rev Sci Instrum
1989
;
60
:
1040
–50.
32
Robinson BH, Mailer C, Reese AW. Linewidth analysis of spin labels in liquids. I. Theory and data analysis.
J Magn Reson
1999
;
138
:
199
–209.
33
Mailer C, Robinson BH, Williams BB, Halpern HJ. Spectral fitting: the extraction of crucial information from a spectrum and a spectral image.
Magn Reson Med
2003
;
49
:
1175
–80.
34
Williams BB, Pan X, Halpern HJ. EPR imaging: the relationship between CW spectra acquired from an extended sample subjected to fixed stepped gradients and the Radon transform of the resonance density.
J Magn Reson
2005
;
174
:
88
–96.
35
Eaton SS, Maltempo MM, Stemp EDA, Eaton GR. Three-dimensional EPR imaging with one spectral and two spatial dimensions.
Chem Phys Lett
1987
;
142
:
567
–9.
36
OxyLite user manual. Oxford Optronix; 2003. p. 56.
37
Collingridge DR, Young WK, Vojnovic B, et al. Measurement of tumor oxygenation: a comparison between polarographic needle electrodes and a time-resolved luminescence-based optical sensor.
Radiat Res
1997
;
147
:
329
–34.
38
Colton T. Statistics in medicine. Boston: Little, Brown &Co.; 1974. p. 372.
39
Bland J, Altman D. Statistical methods for assessing agreement between two methods of clinical measurement.
Lancet
1986
;
1
:
307
–10.
40
Howard-Flanders P. Physical and chemical mechanisms in the injury of cells by ionizing radiation.
Adv Biol Med Phys
1958
;
13
:
553
–603.
41
Mailer C, Sundramoorthy SV, Pelizzari CA, Halpern HJ. Spin echo spectroscopic electron paramagnetic resonance imaging.
Magn Reson Med
2006
;
55
:
904
–12.