Quantitative Tissue Oxygen Measurement in Multiple Organs Using ¹⁹F MRI in a Rat Model

Abstract

Measurement of individual organ tissue oxygen levels can provide information to help evaluate and optimize medical interventions in many areas including wound healing, resuscitation strategies, and cancer therapeutics. Echo planar ¹⁹F MRI has previously focused on tumor oxygen measurement at low oxygen levels (pO₂) < 30 mmHg. It uses the linear relationship between spin-lattice relaxation rate (R₁) of hexafluorobenzene (HFB) and pO₂. The feasibility of this technique for a wider range of pO₂ values and individual organ tissue pO₂ measurement was investigated in a rat model. Spin-lattice relaxation times (T₁=1/R₁) of HFB were measured using ¹⁹F saturation recovery echo planar imaging (EPI). Initial in vitro studies validated the linear relationship between R₁ and pO₂ from 0 mmHg to 760 mmHg oxygen partial pressure at 25°C, 37°C, and 41°C at 7 Tesla for HFB. In vivo experiments measured rat tissue oxygen (ptO₂) levels of brain, kidney, liver, gut, muscle and skin during inhalation of both 30% and 100% oxygen. All organ ptO₂ values significantly increased with hyperoxia (p<0.001). This study demonstrates that ¹⁹F MRI of HFB offers a feasible tool to measure regional ptO₂ in vivo, and that hyperoxia significantly increases ptO₂ of multiple organs in a rat model.

Introduction

The primary goal in treating critically ill patients is to ensure adequate perfusion and oxygenation of vital organs. Tissue oxygenation is complex and depends upon the interaction of blood flow, arterial oxygen tension, hemoglobin dissociation conditions, oxygen carrying capacity, mass transfer resistances and local oxygen consumption. Tissue hypoxia has a high correlation with organ failure and poor prognosis in severe disease states(1,2). Frequently used surrogates for tissue oxygenation include blood pressure, heart rate, cardiac filling pressures, cardiac output, hemoglobin saturation, systemic blood gas measurement, and lactate levels. All are global measures and give no specific information about the status of individual organs. Abnormal values of these surrogate measures often signify a poor prognosis. However, using medical therapies to obtain normal values does not reliably predict a favorable result, and may lead to patient outcomes that are worse, better, or unchanged(3). Measurement of individual organ tissue oxygen levels could help physicians and investigators evaluate and optimize medical interventions in many areas including wound healing(4), resuscitation strategies(5), and cancer therapeutics(6) by providing defined endpoints to titrate therapies and evaluate interventions.

A variety of tissue oxygen measurement techniques exist, but all have limitations. Although polarographic electrodes are considered the gold standard of oximetry, they are invasive, consume oxygen, sample a limited tissue volume, and cannot display the heterogeneity of tissue oxygen levels(7). Although fiberoptic probes do not consume oxygen, they suffer from similar limitations regarding invasiveness and fragility(8). Methods based on hemoglobin (Hb) saturation, such as near infrared spectroscopy, measure an average of arterial, venous, and capillary blood in a large tissue volume. Electron paramagnetic resonance (EPR) gives precise tissue oxygen values and allows repeated measurements, but requires implantation of a paramagnetic material, is limited in its sampling area, and is hindered by a lack of available instrumentation(9). Positron emission tomography images tissue oxygen levels using intravenous reporter molecules that become trapped in the tissues, but the calibration range of this method is limited to hypoxic conditions(10). Blood oxygen level-dependent (BOLD) MRI is sensitive to intravascular Hb saturation but does not provide absolute pO₂. Although the BOLD technique is positively correlated with tumor pO₂, its intravascular based imaging does not reliably predict tissue pO₂(11,12). In addition, it is heavily affected by changes in blood flow and vascular volume(13).

A quantitative oximetry method using fluorine-based MRI has been developed that allows regional dynamic changes to be measured from sequential images of tissue oxygen(14,15). The use of ¹⁹F MRI oximetry based on direct tissue injection of hexafluorobenzene (HFB) has primarily been directed at measurements of tumor oxygenation in animal models, and has consequently focused on pO₂ quantitative imaging at low oxygen levels(16). We have chosen this method because it is minimally invasive, requiring only microliters of a fluorinated substance to be injected, it directly measures pO₂ in the tissue, allows imaging of regional tissue oxygen distributions based on small voxels (typically, in-plane areas of 1.25 mm X 1.25 mm), and has been validated by direct comparison to multiple polarographic electrode measurements(7,15,16) and fluorescence quenching oximetry(17). Various perfluorocarbons (PFCs) have served as reporter molecules, each exhibiting a linear relationship between their corresponding spin lattice relaxation rate (R₁) and pO₂(18). The ability of a PFC to measure pO₂ varies with its molecular properties and route of administration. PFCs injected intravenously tend to be largely sequestered by the reticuloendothelial system into the liver and spleen(19). This method of administration often results in levels of PFC accumulation in other tissues of interest that are too low to provide an adequate signal to noise ratio (SNR). Additionally, intravenous administration has the potential to produce even lower tissue concentrations of PFC in poorly vascularized regions. Even though the technique is not dependent on absolute reporter molecule concentration, these areas would have a SNR too low for accurate pO₂ measurement, and could bias the technique to areas of higher perfusion and oxygenation. In contrast, the technique of direct tissue injection affords greater precision and sensitivity of pO₂ measurements, as well as the ability to study potentially poorly perfused areas(20,21). The PFC hexafluorobenzene (HFB) has 6 magnetically equivalent fluorine atoms, resulting in a single resonance. This feature provides high sensitivity, an advantageous SNR, and dramatically simplifies imaging. HFB also has the advantage of a relatively low sensitivity to both temperature and pH variation, and is easily available through several commercial sources(15,22). HFB has minimal toxicity, with no mutagenicity, teratogenicity, or fetotoxicity reported, and an LD50 > 10g/kg when administered orally or intraperitoneally in rats(20). Thus HFB also has the potential for future use in humans.

Although the vast majority of oxygen content in blood is carried on hemoglobin, increasing inspired oxygen levels results in a significant increase in arterial plasma oxygen content (dissolved oxygen), and is associated with improved wound healing, decreased infection rates, and improved efficacy of cancer treatment(23,24). High levels of inspired oxygen (ranging up to 100%) are used daily in the operating rooms and critical care units to improve arterial oxygenation, although the effects on tissue oxygenation are not normally measured in the clinical setting. Even under normobaric conditions, high oxygen tension in the arterial blood has been reported to lead to hemodynamic changes including bradycardia, decreases in both pulmonary artery pressure and cardiac output, and increases in systemic vascular tone(25,26). Despite many attempts, animal studies have generally failed to identify specific organs or tissue areas of altered blood flow during hyperoxia. An increase in vascular resistance from hyperoxia has been observed in animal models, but blood flow to major organs appeared to be preserved(27–29). The specific mechanisms of both changes in vascular resistance and changes in tissue oxygen consumption remain unknown, leaving the regional effects of hyperoxia unclear, and highlighting the need for direct tissue oxygen measurement.

A better understanding of how hyperoxia affects individual organ oxygenation is needed to optimize the use of oxygen as a medical therapy. Therefore, one of the goals of this study was to examine the influence of increasing the dissolved arterial oxygen content on individual organ pO₂ as measured by ¹⁹F MRI, while ensuring that temperature, arterial CO₂, anesthetic depth, and intravascular volume status remained constant. For this study, we chose inspired oxygen levels that would ensure a nearly complete saturation of hemoglobin, yet provide markedly different arterial blood pO₂. To examine the influence of hyperoxemia on individual organs, we measured brain, gut, kidney, liver, muscle, and skin oxygen tension during ventilation with the fraction of inspired oxygen (FiO₂) set at either 0.3 or 1.0 in a euthermic, euvolemic, anesthetized rat model. An FiO₂ of 0.3 rather than room air (0.21) was chosen because under general anesthesia there is an increase in pulmonary shunting which results in hypoxic arterial oxygen levels compared to awake subjects. In addition, an FiO₂ of less than 0.3 is rarely used in either the intensive care unit (ICU) or operating room for patients that are intubated and anesthetized.

Methods

Magnetic Resonance Pulse Sequence for Measuring the T₁ of HFB

A saturation recovery echo planar imaging (EPI) sequence(15) was used to measure the T₁ of directly injected HFB. A group of twenty 90° hard pulses with 50 ms spacing was applied to saturate the magnetization along the z-axis. Following a variable time delay (τ), a single-shot spin echo with blipped phase encoding EPI was used to collect images. The saturation recovery was favored, and chosen over the traditional inversion recovery since there is no need to wait more than a period of 5T₁ between acquisitions for longitudinal relaxation recovery. This method is well suited for long T₁ measurements, which were common in this study. The Alternated Relaxation Delays with Variable Acquisitions (ARDVARC) scheme (30) was modified and used to acquire 14 τ in the order: 90 s, 200 ms, 60 s, 400 ms, 40 s, 600 ms, 20 s, 800 ms, 16 s, 1 s, 8 s, 1.5 s, 4 s, and 2 s. This acquisition scheme is optimized to minimize any error from the clearance effects of HFB in vivo. A wide range of τ values was selected since the range of the expected T₁ values under differing oxygen conditions is broad. Generally, the corresponding number of averages was set to one for all delays, but occasionally an average of two was needed when the signal-to-noise ratio (SNR) was low. The acquisition time per T₁ measurement was approximately 5 minutes. The standard voxel size was 1.25 × 1.25 × 5 mm.

Calibration of HFB R₁ and pO₂ at 7 Tesla

Calibration techniques were based on previously described methods used at 4.7T magnetic field strength(15,30). Three different HFB samples were equilibrated at each of 5 different oxygen concentrations balanced with nitrogen. The concentrations examined were 0% O₂ (pure N₂), 5% O₂, 10% O₂, 21% O₂, 50% O₂, 100% O₂. These oxygen concentrations were examined under atmospheric pressure (sea level) at 25°C, 37°C, and 41°C. Additionally, since 2% isoflurane was used to anesthetize animals, the entire calibration sample conditions were repeated with and without the presence of 2% isoflurane in order to determine whether the fluorine atoms in isoflurane affect the linear relationship between the R₁ of HFB and pO₂. The pO₂ was calculated using standard partial pressure equations taking into account the barometric pressure, oxygen percentage, temperature, and presence of water vapor(31).

One ml of HFB was added to a gas-tight NMR tube (Wilmad Lab Glass, Buena, NJ, USA) together with 5 ml water. Since HFB is hydrophobic and immiscible with water, the denser HFB remained in a separate layer at the bottom of the tube. The NMR tube containing HFB and water was equilibrated under the desired temperature and atmospheric conditions within a sealed previously equilibrated chamber using a small catheter with the selected gas continuously bubbled into the sample. The desired temperature was maintained during imaging using a water bath containing D₂O circulating in tubing surrounding the NMR tube. The T₁ of HFB in the sample was then determined with ¹⁹F saturation recovery EPI. The experiment was repeated in triplicate for each oxygen and temperature condition of interest. Separate sets of experiments determined the time needed for oxygen equilibration in this experimental set-up and confirmed maintenance of a constant selected temperature during imaging.

Animal preparation

The University of California, San Francisco Institutional Animal Care and Use Committee (IACUC) approved this work and animal protocol prior to any experimentation. Twelve male Sprague-Dawley rats (250–350g, Charles River Labs, Wilmington, MA, USA) were anesthetized with isoflurane. General anesthesia was initially maintained with 2.0% isoflurane in 100% oxygen via nose cone during placement of invasive monitoring and HFB. The rat was shaved, and placed on an adjustable temperature water pad, which was used to keep the animal euthermic (37.5 °C ± 0.5°C). Temperature was measured by rectal probe. An MRI compatible pulse oximeter (Nonin Medical, Inc. Plymouth, MN) was placed on the animal’s right hind paw; heart rate and oxygen saturation were monitored for the duration of the experiment. The trachea was exposed and a tracheal ventilation cannula placed and secured. The rat was then mechanically ventilated throughout the experiment in order to control and maintain normal physiologic arterial CO₂ levels and decrease atelectasis during the experiment. Initial ventilation settings of 70 breaths/min and tidal volume of 12 ml/kg were then adjusted based on end tidal carbon dioxide, which was monitored every 15 minutes throughout the remainder of the surgery and experimental procedure. These higher tidal volume settings were necessary due to the large capacitance in the longer length of tubing needed to position the ventilator outside the magnetic field of the scanner. Following the tracheotomy, the femoral artery was cannulated. A 1.5 cm incision was made along the crease formed by the abdomen and left thigh (contralateral to the side with the pulse oximeter). The femoral artery was then exposed. Approximately 1 cm of the vessel was dissected free under microscopic visualization and a saline-filled catheter inserted into the vessel and secured in place with suture. This arterial line was then connected to a blood pressure monitor (Spacelabs Healthcare, Issaquah, WA, USA) and used throughout the remainder of the preparation and data collection for blood pressure monitoring and blood gas sampling. After securing the arterial line, the tail vein was cannulated with a 24 gauge catheter and attached to a programmable syringe infusion pump (New Era Pump Systems, Inc., Wantagh, NY, USA); 10mL/kg normal saline was bolused at this time to replace losses during surgery, and then normal saline infused intravenously at a rate of 2.5mL/kg/hr for fluid maintenance. Total surgical time ranged from 60 to 120 minutes.

The animal was placed on its side, and a superficial skin nick (2mm) was made on the proximal hind thigh, contralateral to the arterial cannulation. Aliquots of HFB were injected into multiple tissues of each rat as follows: 50 µL HFB was incrementally injected under direct vision into the muscle with a 33 ga. needle. Percutaneous injection was used for placement of 50 µL HFB into the skin. Following a small scalp incision, a 1.5 mm diameter hole was drilled through the animal’s skull, 2mm anterior to the bregma and 4mm lateral from midline. 30 µL HFB was injected 1.5 mm below the dural layer over 10 minutes with the assistance of a stereotactic frame and microscope. A 1.5 to 2.0 cm incision was made just caudal to the right inferior costal margin, through the peritoneum to expose the kidney, liver, and gut. 50 µL HFB was injected into the liver tissue, kidney tissue, and gut wall under direct vision. Total time for HFB injection for all organs of interest was approximately 60 minutes.

MRI Experiments

All experiments were performed on a Varian DirectDrive™ 7T horizontal bore magnet equipped with 21-cm inner diameter imaging gradients that provide ±200 mT/m (Varian, Inc., Palo Alto, CA, USA). A custom constructed ¹⁹F/¹H dually tunable birdcage RF transmit/receive coil (Rapid Biomedical GmbH, Rimpar, Germany) with 7.2 cm inner diameter was used for proton and fluorine imaging. Each animal was randomly assigned to an initial FiO₂ of 0.3 or 1.0. Following surgical preparation and HFB injection, the animal was allowed to equilibrate on the assigned FiO₂ for 20 minutes, and an arterial blood gas (ABG; 0.25 ml) was sampled and analyzed using an i-STAT® hand held blood analyzer (Abbott Point of Care, Inc., East Windsor, NJ, USA). The rat was kept prone on a D₂O circulating warming pad and covered with a plastic covered cotton drape throughout the experimentation to maintain euthermia. The animal and monitoring lines were secured onto a custom built Plexiglas bed and cradle inserted into the magnet such that the RF coil center was aligned to the isocenter of the imaging gradients. The bed was manually adjusted to position each different organ location at the center of the RF coil. The order of imaging each organ of interest was randomized. Conventional T₁-weighted 2D spin echo ¹H and ¹⁹F images with typical voxel size of 0.625×0.625×2 mm were acquired to identify the position of HFB signal for different organs of interest. Following the anatomical scans, the T₁ of the implanted HFB was measured with the saturation recovery EPI. The field of view (FOV) of the saturation recovery EPI images was the same as the spin echo images for appropriate signal registration.

Ventilation, administration of isoflurane, and continuous monitoring of pulse and blood pressure were continued throughout the MRI scan, with all ferrous metal monitors and equipment located outside the 5-gauss field line. Rectal temperature was recorded prior to placement of the rat in the magnet and again following completion of the scan. Abdominal images were acquired with respiratory gating. The respiration signal was determined using a mechanical pressure transduction signal from the ventilator exhaust outlet that was transferred into a respiratory gating module (Small Animal Instruments, Inc, Stony Brook, NY, USA).

Following the organ scans at the initial FiO₂, the inspired gas was changed to the alternate value. The animal was rescanned as described above after a 20-minute equilibration period. Following the scan, 0.25 ml of arterial blood was again taken for blood gas and hematocrit analysis. The total scan time for both FiO₂ values was approximately 2.5 hrs.

After completion of imaging, the animal was removed from the magnet, a final blood gas drawn, and the animal euthanized.

Data Analysis

Raw data were imported into Matlab (MathWorks Inc., Natick, MA, USA) for image reconstruction and processing. The T₁ was calculated voxel by voxel in the selected region of interest (ROI) with a three-parameter fit using the Levenberg-Marquardt algorithm(32). Error estimates of the derived T₁ were calculated from the residuals and Jacobian matrix used in this fit method. The in vitro calibration data at each temperature were then fit to Eq. [1] as previously detailed by Zhao et al.(15):

$${\mathrm{R}}_1=\mathrm{A}+\mathrm{B}*{\text{psO}}_2$$ [1]

Where R₁ is the spin lattice relaxation rate (s⁻¹), psO₂ is the oxygen pressure in the sample (mmHg), A and B are constants at a given temperature and magnetic field. R₁ was derived from the reciprocal of the mean T₁ averaged over selected ROIs in three individual measurements.

Prior research has demonstrated a linear dependence of R₁ on temperature over a small range(33). Assuming a linear dependence of A and B with temperature T, then A= C+D*T and B= E+ F*T, and Eq. [1] is transferred into a temperature-dependent model described by Eq. [2]:

$${\mathrm{R}}_1=\mathrm{C}+\mathrm{D}*\mathrm{T}+(\mathrm{E}+\mathrm{F}*\mathrm{T})*{\text{psO}}_2$$ [2]

where C, D, E, F are constants, T is the temperature in the unit of Celsius. These results were used to examine the effects of temperature at a field strength of 7T.

The in vivo tissue oxygen pressure (ptO₂) was calculated for each voxel in the selected ROI using the linear calibration equation derived at 37°C shown in Eq. [3] in the Results section. To improve data quality, similar to previous analysis methods(15), any voxels with calculated T₁ error greater than 40% (determined from the curve fit to derive T₁) were omitted from statistical analysis. In addition, any extreme outliers were also removed by combining voxels from all animals for a given condition, and removing voxels with ptO₂ outside the mean ptO₂ ± 2.5SD. Application of these two criteria resulted in removal of 169 voxels and 21 voxels respectively from an original total of 1190.

Statistical Analysis

Baseline physiologic rat data were compared between an FiO₂ of 0.3 and 1.0 using a paired t-test. A linear mixed effects model was used for statistical data analysis of ptO₂ as the primary study outcome, with condition (FiO₂ of 0.3 or 1.0) and organ as the two primary fixed effects. The entire data set was analyzed using a two-way repeated measures analysis of variance (ANOVA). Because we were specifically interested in different effects by organ, we tested all pair-wise differences in the regression-adjusted ("least squares") estimates of organ differences, using the post hoc Tukey-Kramer test to adjust for multiple comparisons. In addition, separate one-way ANOVA’s were completed across organs for each inhaled oxygen condition in order to alleviate the reduced power of the numerous multiple comparisons of the two-way ANOVA. All calculations were done using SAS/STAT software (Version 9.1.3 of the SAS System for Windows Copyright ® 2002–2003 and JMP 7.01, SAS Institute Inc., Cary, NC, USA). Under this study design, with a planned sample size of 12 and the anticipated effect size based on previously published values of rat organ ptO₂(9,34,35), our a priori power analysis determined that we should have > 85% power to detect a statistically significant difference with an alpha level set at 0.05 under the null hypothesis of no organ site difference.

Results

Calibration Results

No significant differences were found in sample calibration results (Table 1) with and without 2% isoflurane for a given temperature. Therefore only results of samples with the addition of 2% isoflurane are shown in Figure 1 to simulate the in vivo environment. The T₁ of HFB ranged from 9.8 s at 0% oxygen to 0.77 s at 100% oxygen at 37°C. The R₁ of HFB was linearly dependent on pO₂ from 0% oxygen to 100% oxygen at 25°C, 37°C and 41°C, shown in Figure 2. At 37°C, A = 0.1041 ± 0.0014 s⁻¹, B = 0.001659 ± 0.000004 (s mmHg)⁻¹ were derived with linear fit (R² = 0.99998) according to Eq. [1]. Thereby the calibration equation at 37°C can be written as follows:

$${\text{pO}}_2(\text{mmHg})=({\mathrm{R}}_1({\mathrm{s}}^{-1})-0.1041)/0.001659$$ [3]

As shown in Figure 1, the HFB R₁ temperature dependence increased as pO₂ increased. Fitting the R₁ of HFB at different pO₂’s and temperatures to the temperature-dependent model Eq. [2] gave C = 0.1500 ± 0.0040 s⁻¹, D = −(1.21 ± 0.12) × 10⁻³ (s °C)⁻¹, E = (2.69 ± 0.06) × 10⁻³ (s mmHg)⁻¹, and F = −(2.74 ± 0.18) × 10⁻⁵ (s mmHg °C)⁻¹.

Table 1.

The intercept and slope obtained from calibration data at 0% or 2% isoflurane at three temperatures

Isoflurane (%)	Temperature(°C)	Intercept (s−1)	Slope 10−3 (s mmHg)−1
0	25	0.121 ± 0.004	1.997 ± 0.013
0	37	0.102 ± 0.003	1.658 ± 0.008
0	41	0.099 ± 0.003	1.576 ± 0.010
2	25	0.120 ± 0.004	2.008 ± 0.012
2	37	0.104 ± 0.001	1.659 ± 0.004
2	41	0.101 ± 0.002	1.578 ± 0.006

Data are mean ± SD from three individual samples using the linear fit from Eq. [1]. No significant effect on the intercept or slope from the presence of isoflurane.

Fig. 1.

Linear fits of R₁ of HFB (with 2% Isoflurane) vs. pO₂ from 0% to 100% at 25°C, 37°C and 41°C. Each sample point represents the mean of 3 individual samples (SD < 0.0153 s⁻¹ at all points displayed).

Fig. 2.

The dependence of intercept and slope on temperature as depicted from Table 1. Data from calibration with and without the presence of isoflurane are displayed. a: Linear fits of intercept vs. temperature according to Eq. [2]. b: Linear fits of slope vs. temperature according to Eq. [2].

Figure 2 displays the calibration curve temperature dependence of both the intercept (2a) and slope (2b) with and without the presence of isoflurane.

In vivo Results

The mean rat physiologic data from the in vivo experimentation are provided in Table 2. Mean arterial pO₂ was 98 ± 12 mmHg at an FiO₂ of 0.3 and 400 ± 85 mmHg at an FiO₂ of 1.0, reflecting the specific significant increase of dissolved oxygen in plasma due to the increase of FiO₂ (p<.01). No significant differences in the remaining measured physiologic data were noted. Mean arterial CO₂ was 38 ± 5 mmHg, mean Hb was 12.8 ± 0.9 g/dL. Both parameters remained in the physiological range throughout the in vivo experiments. Mean weight of 12 rats was 312 ± 39 grams. Core temperature of animals remained euthermic ranging from 37.0°C to 38.0°C during all experiments.

Table 2.

Physiological values measured during in vivo experimentation at FiO₂ of 0.3 and 1.0

Physiological Variables	FiO2 = 0.3	FiO2 = 1.0
Arterial pO2 (mmHg) *	98 ± 12	400 ± 85
Arterial pCO2 (mmHg)	36 ± 5	40 ± 4
Hb (g/dL)	12.8 ± 0.8	12.9 ± 0.9
Mean arterial pressure(mmHg)	90 ± 6	86 ± 9
Temperature (°C)	37.3 ± 0.3	37.3 ± 0.6
Heart rate (beats/minute)	352 ± 38	353 ± 23

Data are mean ± SD. A significant difference (*) was noted in the arterial pO₂ (p<.001) between the conditions. No significant difference (p<.05) was noted between conditions for the remaining physiologic variables.

The distributions of ptO₂ in all six organs of interest at FiO₂ conditions of 0.3 and 1.0 are shown in Figure 3. These ¹⁹F MRI based images are from a single representative animal. The shift towards yellow and orange colors displayed in the ptO₂ mapping on the right side represents higher values ptO₂ in all scanned organs at FiO₂ of 1.0 compared to 0.3. Separate organ histograms of the ptO₂ values recorded from all animals are displayed in Figure 4. Tissue oxygenation is heterogeneous in individual organs and is noted to vary significantly between organs at the higher level of inspired oxygen (FiO2 = 1.0).

Fig. 3.

Tissue oxygen distributions of six organs at FiO₂ of 0.3 (left) and 1.0 (right) shown in color superimposed on proton anatomical images shown in black and white. PO₂ reference ranges corresponding to different colors are shown on the side bar color key.

Fig. 4.

Separate organ histograms showing the distribution of individual voxel ptO₂ values recorded from all 12 animals are displayed at both an FiO₂ of 0.3 (gray) and 1.0 (black).

Mean values of ptO₂ were calculated over all voxels in the selected ROI for each organ in each individual animal. The mean organ ptO₂ from all animals under each inspired oxygen condition are displayed graphically in Figure 5. The mean organ ptO₂ values at an FiO₂ of 0.3 ranged from 57 ±13 mmHg to 75 ± 15 mmHg. The mean organ ptO₂ values at an FiO₂ of 1.0 ranged from 127 ± 44 mmHg to 197 ± 86 mmHg. Our final mixed model (two-way ANOVA) showed both "condition" (p-value <.001) and "organ" (p-value <.001) to be significant predictors of ptO₂ level. For each organ studied, the ptO₂ measured with an FiO₂ = 1.0 was significantly greater than that measured while the animal was ventilated with an FiO₂ = 0.3 (p < .05). This result demonstrates a significant positive increase in organ ptO₂ values at the higher inspired oxygen condition. Comparison of organs showed that the subgroup ptO₂ values of skin and muscle were significantly greater than brain and gut which in turn were significantly greater than kidney and liver ptO₂ values (p<.05). A second analysis across organs using a one-way ANOVA at each FiO₂ condition found a significant organ ptO₂ difference noted with an FiO₂ = 1.0 (similar to the two-way ANOVA), but no significant difference between organ ptO₂ values at an FiO₂ =0.3.

Fig. 5.

Bar plots of mean ptO₂ over all voxels in brain, kidney, liver, gut, muscle and skin among 12 studied animals at FiO₂ of 0.3 and 1.0. Significant differences (*) are noted for all organs as the ptO₂ at an FiO₂ of 1.0 is significantly increased compared to values at an FiO₂ of 0.3 (p<0.05).

DISCUSSION

We demonstrated the feasibility of quantitative tissue oxygen mapping in a rat model with ¹⁹F MRI measurement of directly injected HFB in multiple organs under two inspired oxygen conditions. Calibration results showed that the spin-lattice relaxation rate R₁ of HFB is linearly dependent on oxygen pressure pO₂ at a fixed temperature in the physiological range 25–41°C from 0% to 100% oxygen. Calibration Eq. [3] is similar to the calibration equation previously derived from 0% to 21% oxygen at 7T(22). The R₁ of HFB is highly sensitive to pO₂, but minimally sensitive to temperature. Since the rat core temperature was carefully maintained between 37°C and 38°C throughout in vivo experimentation, the error of ptO₂ measurement was minimized (6 mmHg/°C at pO₂ 300 mmHg). According to calibration results, 2% isoflurane used for anesthesia in vivo has no significant effects on the calibration equation, and consequently inhaled perfluorocarbon anesthetics can be used without worry of affecting the results of a ¹⁹F MRI study using HFB. Prior studies demonstrated that the linear dependence of R₁ of PFCs on pO₂ is also essentially unresponsive to pH, CO₂, charged paramagnetic ions, mixing with blood, or emulsification(15,36). This allows ¹⁹F MRI to remain a viable option for measuring ptO₂ under a variety of physiologic conditions.

The mean values of ptO₂ using ¹⁹F MRI in this study are consistently somewhat higher than literature values previously reported for corresponding measurements of brain, kidney, and muscle in rats measured by EPR(9,34,35). The use of EPR ptO₂ measurement shows a similar significant rise in ptO₂ with increasing inspired oxygen concentration in measurement of both rat and mouse specific organs(9,34,37). The ¹⁹F MRI oximetry values are also somewhat higher than previous rat organ measurements of ptO₂ obtained using polarographic electrodes(38–42). In a previous study by Jordan et al. using both ¹⁹F MRI oximetry and fluorescence quenching fiber-optic probe oximetry to quantify the in vivo response of mouse tumor pO₂ with inhalation of carbogen, the tumor pO₂ determined with the ¹⁹F MRI technique was typically higher(17). The exact reason for the higher absolute values with organ ptO₂ measurement using ¹⁹F MRI compared to EPR, polarographic electrode, and fluorescence quenching measurements remains uncertain, but may be due to differences in tissue disruption, use of a mean of multiple measurement voxels (with ¹⁹F MRI) vs. a single value averaging a larger area of contacting tissue, length of time between placement of reporter material and measurement, anesthetic technique, and control of specific variables known to change tissue perfusion such as temperature, volume status, and spontaneous vs. controlled ventilation. We observed that HFB signal was often imaged near the ventricular floor of the brain instead of cortex layer at the top in Fig. 3. We suspect that the HFB entered the cerebral spinal fluid (CSF) in either ventricles or subdural space in many studies despite a careful slow injection aided by a stereotactic syringe holder and microscope. The parenchymal injection is difficult to reliably reproduce as this layer in the rat brain is approximately 1 – 2 mm thick and HFB will enter the CSF if the needle tip is either to shallow or deep, frequently traveling to the ventricle and sinking to the floor due to its hydrophobic nature and greater density than water. The failure of HFB injection in the parenchyma may overestimate brain tissue oxygen levels since pO₂ in CSF is often higher than in brain tissue itself(43).

The marked variability and different absolute increases in the organ ptO₂ noted in our study using ¹⁹F MRI, show the diversities of oxygenation in various organs and the potential benefit of the regional versus global measurements this technique allows. The increased amount of oxygen bound to hemoglobin due to increasing the FiO₂ from 0.3 to 1.0 is small since both levels of corresponding arterial pO₂ (98 and 400 mmHg respectively) represent conditions in which hemoglobin is nearly fully saturated. However the mean arterial pO₂ at an FiO₂ of 1.0 was four fold higher than values corresponding to an FiO₂ of 0.3, which reflects the large increase of oxygen dissolved in plasma. Therefore, we conclude that the increases in ptO₂ quantified in all studied organs were primarily caused by the increased dissolved oxygen level in plasma. Although the increase in arterial blood oxygen content due to the increased FiO₂ is small when compared to the total oxygen content, the corresponding increase of arterial pO₂ tension significantly increases the driving force of oxygen diffusion from the capillary bed into tissue, and thus increases the absolute organ ptO₂. We believe the ¹⁹F MRI technique does not reflect an unintended direct measure of the arterial pO₂, as all organ tissue values are significantly less than the arterial oxygen levels, and any HFB in direct contact with the vascular system would be rapidly removed from the tissue, taken to the lungs and exhaled.

Similar absolute increases of ptO₂ were found regardless of the order of inspired oxygen levels. This indicates that the tissue damage from the HFB injection was likely minimal. However, during the injection of HFB, transient reductions of blood pressure were observed occasionally. HFB has been used in the past as an inhaled anesthetic(44). These brief hypotensive episodes were most likely caused from a small portion of the HFB entering the intravascular compartment during the injection and exerting an anesthetic effect and its associated brief period of vasodilation or decreased myocardial contractility associated with most anesthetics. In order to prevent the transient acute blood pressure drop, we slowed the injection rate in each organ and reduced the isoflurane concentration briefly during the HFB injection period. The observed retention period of HFB was different in each organ. We believe that the residence time and clearance of HFB is most likely dependent on the individual organ perfusion. Half lives of HFB in less perfused organs such as muscle and skin are reported to be approximately 10 hours(18). HFB signal decayed faster in the more vascularized organs, eg. liver and kidney. The alternated relaxation delays were used to minimize the error in T₁ measurement that might otherwise result from the clearance effect of HFB. Two signal averages were used to compensate for the signal loss when a low SNR of HFB was a concern for MRI scans at later post-injection times.

In summary, this minimally invasive ¹⁹F MRI oximetry technique using HFB as a reporter probe provides a sensitive quantitative method for measuring regional organ tissue oxygen tension and dynamic changes. These measures represent the balance between oxygen delivery and metabolic use, providing an endpoint to help guide future medical research and optimize a variety of therapeutic interventions, resuscitation strategies, and disease processes.