Rapid quantification of oxygen tension in blood flow with a fluorine nanoparticle reporter and a novel Blood flow-Enhanced-Saturation-Recovery (BESR) sequence

Abstract

We present a novel Blood flow-Enhanced-Saturation-Recovery (BESR) sequence, which allows rapid in vivo T1 measurement of blood for both ¹H and ¹⁹F nuclei. BESR sequence is achieved by combining homogeneous spin preparation and time-of-flight image acquisition and therefore preserves high time efficiency and SNR for ¹⁹F imaging of circulating Perfluorocarbon (PFC) Nanoparticles (NPs) comprising a perfluoro-15-crown-5-ether core and a lipid monolayer (nominal size = 250 nm). The consistency and accuracy of the BESR sequence for measuring T1 of blood was validated experimentally. With a confirmed linear response feature of ¹⁹F R1 with oxygen tension in both salt solution and blood sample, we demonstrated the feasibility of the BESR sequence to quantitatively determine the oxygen tension within mouse left and right ventricles under both normoxia and hyperoxia conditions. Thus, ¹⁹F BESR MRI of circulating PFC NPs represents a new approach to non-invasively evaluate intravascular oxygen tension.

Introduction

Blood oxygen tension (pO₂) is an essential pathophysiological index of circulatory and respiratory health that is useful for characterizing various disorders such as congenital heart disease (1), arteriovenous malformations (2), atherosclerotic dysfunction (3), and overt or occult heart failure in general (4). The current gold standard for pO₂ measurement entails direct blood sampling using arterial puncture, or invasive catheterization when A-V O₂ differences, or cardiac output and oxygen consumption data are needed.

MRI has been exploited as an alternative tool for non-invasive assessment of blood oxygenation based on blood-oxygenation-level-dependant (BOLD) MRI determined ¹H R2* (5), susceptibility (6), or T2 and T1 contrast (7-8). Among these methods, R2* and susceptibility contrast generated by deoxygenated-hemoglobin (deoxy-Hb) can be used to estimate deoxy-Hb concentration and blood saturation (SO₂) because deoxy-Hb is highly paramagnetic. Nevertheless, the readout of BOLD and susceptibility-dependent contrast is qualitative, or otherwise requires organ-specific prior knowledge to build complex mathematical models (5-6). Blood flow imaging with global T2 preparation has also been employed to measure blood oxygenation because of its linear correlation with blood ¹H R2. However, for in vivo studies, B₁ and B₀ inhomogeneity and pulsatile flow artifacts pose major obstacles for effective spin preparation and consistent R2 estimation (8). Additionally, blood ¹H R1 is also linearly responsive to dissolved pO₂ and thus may also be employed as a quantitative marker for blood oxygenation. However, because of the limited solubility of O₂ in plasma, the slope of ¹H R1 as a function of pO₂ is as low as 2×10^-4s^-1/mmHg, resulting in significant uncertainty for in vivo pO₂ quantification (7).

Perfluorocarbon nanoparticles (PFC NPs) have been proposed as an ¹⁹F MR pO₂ sensor (9-10) as a consequence of two attributes: (a) O₂ is a weakly paramagnetic molecule with unpaired electrons and net magnetization; (b) O₂ exhibits high solubility in PFC; (c) the linear O₂ release response curve of PFC and the free O₂ diffusion between PFC and surrounding environment render PFC as a reliable pO₂ probe for in vivo applications of measuring vascular pO₂ (11-12) and tissue oxygenation (9,13-14). Previous reports have demonstrated that the slope of ¹⁹F R1 of PFC as a function of pO₂ could be ten-fold greater than the dependence of ¹H R1 on pO₂ (12-13,15). Additionally, because the lipid monolayer of PFC NPs is impermeable to macromolecules and ions, free paramagnetic macromolecules and ions in the blood stream do not affect the measurement of ¹⁹F R1 emanating from PFC NPs, or its dependency on pO₂ (16). Finally, as a FDA-approved oxygen delivery vehicle, PFC NPs exhibit an acceptable bio-safety profile for ultimate clinical use (17-18).

Similar to ¹H MRI (19-22), one major obstacle of intravascular pO₂ measurements based on ¹⁹F T1 of PFC NPs is the blood pulsatile motion artifact and in-flow effect. Other limitations include relatively low signal-to-noise ratio (SNR) due to limited amount of injected PFC NPs in the blood pool, and the inhomogeneity of spin preparation due to the movement of arterial blood through a large portion of body during the relaxation measurement. Therefore, the implementation of novel ¹⁹F MRI strategies to minimize these artifacts would be critical to blood pO₂ measurements obtained by ¹⁹F MRI relaxometry.

In the present work, we have developed a novel Blood flow-Enhanced-Saturation-Recovery (BESR) sequence for measuring ¹⁹F T1 of flowing PFC NPs in blood pool. The BESR sequence preserves the time efficiency of the classical Look-Locker sequence and the high SNR of saturation recovery sequence. We first performed a computer simulation to illustrate the homogeneity of spin preparation in the BESR sequence, and to define the effect of limited SNR on the precision of the pO₂ measurement. The accuracy of BESR sequence measured T1 was tested on a phantom and evaluated in vivo by measuring arterial ¹H T1 in carotid arteries and cardiac ventricles. Finally, we optimized scan times and PFC NPs doses in conjunction with the BESR sequence for ¹⁹F imaging of circulating PFC NPs to provide a quantitative estimate of arterial and venous pO₂ in mouse heart ventricles under both normoxia and hyperoxia conditions.

Methods

The BESR sequence was implemented and all images were acquired on a 4.7 T Varian INNOVA scanner. All procedures in this study conformed to the guidelines and approvals by the Animal Studies Committee of Washington University in St. Louis.

Pulse Sequence and Imaging Setup

The BESR pulse sequence is designed for measurement of pO₂ in pulsatile blood flow. It comprises two parts: homogeneous spin preparation and time-of-flight image acquisition (Fig. 1). The spin preparation is achieved with a non-selective whole body saturation pulse train and the image acquisition is realized with gradient echoes at high repetition rate. Both the pre-pulses and imaging excitation pulses are gated to the ECG with appropriate trigger delay to minimize the motion artifact associating with the beating heart and vascular contraction and dilation.

Figure 1.

Illustration of the Blood flow-Enhanced-Saturation-Recovery (BESR) sequence. Spin preparation is achieved with multiple non-selective saturation pulses; image acquisition uses a gradient echo approach with cardiac triggering; saturation pre-pulse train and gradient echo imaging are both triggered by ECG. Each image acquisition is acquired at the same trigger delay time t but corresponds to different saturation delay time TS.

In contrast to measuring the T1 of solid tissues, the spin preparation for measuring T1 of flowing blood should be non-selective and homogenous over the entire subject because spins in the blood that will be excited for imaging might arise from a variety of locations during the spin preparation step (22-23). To achieve homogeneous spin preparation, an actively-decoupled transmit-only saddle coil (O.D. = 7cm, length = 12 cm) that covers the entire mouse body and a composite RF saturation pulse train comprising five constitutive 90° sinc-shaped pulses are employed. The bandwidth of each saturation pulse is 12 KHz and the time delay between two adjacent saturation pulses is 2 ms, during which crusher gradients along different directions are applied. The homogeneity of spin preparation is expected to improve progressively with the number of 90° pulses in the saturation pulse train. Specifically, s(x, y, z), the percentage residual magnetization along z direction in a voxel positioned at (x, y, z) after n closely spaced saturation pulses followed by crusher gradients can be written as

$$s(x,y,z)={[\cos (\phi (x,y,z))]}^n,$$ [1]

where φ(x, y, z) is the flip angle at (x, y, z) and is determined by the B₁ field profile according to φ(x, y, z) = B₁(x, y, z)τγ, where γ is the Larmor frequency and τ is the pulse width. To confirm the homogenous saturation over the region of interest (ROI), we first acquired a flip angle mapping of our transmit coil on a water phantom (D = 6 cm, length = 10 cm) using a dual angle B1 mapping approach with α = 30° (24). Quantitative residual magnetization mapping after various number of saturation pulses (NSP) was then obtained by adjusting NSP in the BESR sequence.

In traditional ¹H saturation recovery T1 mapping, only one imaging excitation pulse is allowed within in a single TR to avoid partial saturation effects arising from frequent repetitive excitations (25). In contrast, because of relatively fast blood flow in major blood vessels (>1 cm/s) and the “time-of-flight” effect, we anticipate that excited spins in the previous echo will flow out of the imaging slice as long as the time interval between two temporally adjacent excitation pulses is sufficiently long (e.g., >0.2 s for 2 mm slice). Within a single TR of the BESR sequence, multiple 90° excitations, each followed by a same phase encoding gradient, are applied for gradient echo imaging. Thus, all echoes acquired in a single TR correspond to the same k-space line at different saturation delays (TS). All the imaging excitation pulses are synchronized with the ECG to minimize the effect of heart rate variations during imaging (26). Finally, to achieve an optimal SNR with BESR sequence, a dedicated actively-decoupled receive-only surface coil (diameter ≈ 3 cm) is placed close to the region of interest for signal reception.

Estimating the effect of SNR on the accuracy of pO₂ measurement

The uncertainty of the pO₂ estimation caused by random noise in saturation recovery imaging was evaluated using computer simulations in Matlab (Mathworks, USA) and conventional saturation recovery imaging on PFC NPs phantoms. A region of interest (ROI) containing 100 pixels was defined to simulate the number of pixels in heart ventricles. Normally distributed Gaussian noise was imposed on the ¹⁹F signal intensity in each pixel to achieve variable SNR. The average noise-contaminated signal in entire ROI was used for non-linear fitting of T1 values. The conversion between T1 and pO₂ values was based on the pre-acquired calibration curve. In the simulation, the following saturation delays were used: TS = 0.25, 0.5, 0.75, 1, 1.25, 1.5, 1.75, 2 and 10 s. Simulation was repeated 10000 times and the standard deviation of the fitted pO₂ was calculated to represent the measurement uncertainty propagated from imaging noise through non-linear fitting. For phantom imaging, three PFC NPs phantoms with volume concentration of 5%, 10% and 20% were sequentially bubbled with N₂, room air and O₂ for 10 minutes. The phantom temperature was maintained at 37°C with circulating water bath. A saturation recovery fast spin echo sequence was used for T1 mapping: TR=3 s, FOV=25 cm × 25 cm, TE = 8.6 ms, echo train length = 16, Matrix size=128 × 128. The standard deviation of the fitted pO₂ values for each single pixel was used to calculate the uncertainty of pO₂ estimate in an ROI containing 100 pixels.

In vitro and in vivo validation of BESR sequence

The accuracy and precision of the BESR sequence was tested by measuring ¹H T1 of circulating water in a tube phantom, which simulated the in vivo setting of ¹⁹F MRI with all signal contributed only by intravascular PFC NPs devoid of tissue background. Briefly, approximately 80 cm silicon tubing (1/8” inner diameter) was twisted and coiled to form an 8-cm long phantom with multiple loops. The phantom was then connected to a pump to generate constant 5 cm/s flow velocity to mimic blood flow. The special design of the phantom together with the carefully chosen flow velocity simulate the blood flow situation in vivo, such that the flowing water could experience a homogenous spin preparation at the beginning of each TR (5 s), but is moving rapidly enough to generate the desired “time-of-flight” effect to replenish the imaging slice with unexcited spins before each gradient-echo (10 echoes in each TR ranging from 0.1 s to 4.6 s time interval between consecutive imaging acquisitions = 0.5 s). Upon the completion of BESR measurement, the pump was turned off and the ¹H T1 of stationary water was measured with a saturation recovery fast spin echo sequence. The image parameters of fast spin echo were TR=5 s, TE = 8.6 ms, echo train length =16, TS=0.1 to 4.6 s with intervals of 0.5 s.

To confirm that the BESR sequence can achieve reproducible T1 measurements in vivo, ¹H T1 measurements of flowing arterial blood in carotid arteries and heart ventricles were acquired in Swiss Webster mice (n=5) at 4.7 T. Mice were anesthetized through inhalation of 1.5% isoflurane ventilated by pure O₂ (2 L/min). Body temperature was maintained at 37°C and ECG was monitored with a small animal monitor system (SAI Inc., USA). During MRI, the isoflurane level was adjusted to maintain mouse heart rate at around 350 bpm. Imaging parameters were: TR=2.5 s, TE = 2.8 ms, FOV=25 cm×25 cm, matrix size=96×96 and number of TS=7. After in vivo imaging, mice were sacrificed and blood was collected from LV. T1 of the collected stationary blood was measured using a custom-built solenoid coil and traditional inversion recovery spectroscopy at 37°C after saturating with pure O₂.

PFC NP formulation and in vitro T1 calibration

Perfluoro-15-crown-5-ether (CE) nanoparticles were formulated as previously described. Briefly, the CE emulsion was composed of 40% (v/v) of CE (Exfluor Research Corp., USA), 2.0% (w/v) of a surfactant commixture, and 1.7% (w/v) glycerin, with water comprising the balance. In vitro calibration of ¹⁹F R1 as a function of pO₂ was performed at three different O₂ concentrations (0%, 21%, and 100% O₂ balanced with N₂) by bubbling gas mixture into a CE emulsion sample for 30–60 min at both room temperature and 37°C. ¹⁹F T1 measurements were made using inversion recovery spectroscopy with TR=10 s, and 10 inversion delays (TI) ranging from 3 ms to 5 s. To confirm that the linear response curve of CE NP on pO₂ was preserved in the blood, mouse blood (Bioreclamation, LLC, USA) containing 10% volume/volume ratio of CE NP was placed in a gas chamber to allow gas exchange through semi-permeable silicon tubing for 30 minutes. Blood was transferred to a sealed vial and the sample temperature was maintained at 37°C. For the linear fitting, the conversion of partial percentage of oxygen to mmHg was based on the assumption that 100% oxygen is equivalent to 713 mmHg (considering the vapor pressure at 37°C is 47 mmHg).

Noninvasive intravascular pO₂ measurement with BESR ¹⁹F imaging

Swiss Webster mice (n=5) were anesthetized with 1-2% isoflurane mixed with room air or O₂ followed by intravenous injection of CE NP (5 ml/kg) via the tail vein. The ¹⁹F signal from isoflurane was barely detectable by MRI in contrast to the strong ¹⁹F single emanating from CE NP in the blood stream, which is likely due to low blood/tissue retention and the short T2* of isoflurane (27). ¹H gradient-echo cine images were acquired initially to position the slices to cover both left and right ventricles (LV and RV). Blood ¹⁹F T1 measurements were carried out under separate conditions of room air and then pure oxygen ventilation. The saddle coil and the surface coil were re-tuned as needed to the proper frequency of 201.5 MHz and 189.6 MHz for ¹H and ¹⁹F imaging, respectively. Parameters for ¹⁹F BESR T1 measurement were: TR, 2.5 s; TE, 2.2 ms; number of points along recovery curve, 8; number of averages, 16; in plane resolution, 1.5 mm × 0.75 mm; slice thickness, 2 mm. In order to optimize non-linear fitting for ¹⁹F T1 of less-oxygenated venous blood exhibiting T1 > 2 s, another set of BESR images acquired with identical imaging parameters but without pre-saturation pulses were used as fully recovered signal. The previously measured ¹⁹F R1 vs pO₂ calibration curve was used for quantifying intravascular pO₂ in vivo.

Data analysis and Statistics

The left and right ventricles and vessel lumen were manually segmented on ¹H and ¹⁹F images. ROI was defined as the region exhibiting > 3 SD higher ¹⁹F signal intensity, i.e. signal contributed by PFC NPs in the blood, as compared with that of surrounding tissues. The average intensity in each ROI was used for fitting data to a 3-parameter saturation recovery curve:

$$S(\mathit{TS})=A-B{e}^{-\mathit{TS}/T_1}.$$ [2]

All statistical analysis was performed using Origin software (OriginLab, USA). Metrics for precision and accuracy of T1 measurements with the BESR sequence were assessed based on ¹H image of the tubing phantom. The precision error was calculated as the root-mean-square of 5 individual measurement normalized by the mean value, and the accuracy error was considered to be the relative difference between the T1 mean value measured using a reference method and the BESR sequence. A two sample paired t-test was used for statistical analysis to test the difference between BESR measured ¹H T1 values and reference ¹H T1 values in vivo, and the difference of ¹⁹F MRI determined pO₂ between hyperoxic and normoxic conditions. A significance threshold of p < 0.05 was used for hypothesis testing.

Results

Mesurement of the homogeneity of pulse train saturation

The flip angle mapping of the transmit coil and the percentage residual magnetization intensity after different number of saturation pulses (1, 3, and 5) were measured experimentally and shown in Fig. 2. As the number of saturation pulses increased, spin nulling in the whole FOV became more uniform despite the inhomogeneity of the B1 profile. After five consecutive 90° pulses, the residual magnetization within the ROI where mice were placed for in vivo experiments (border marked by yellow dash line) was < 4%, substantially lower than the > 20% residual magnetization after a single saturation pulse.

Figure 2.

(a, b) Flip angle mapping of a water phantom in axial and coronal plane. (c, d) Measured residual magnetization after different number of saturation pulses (NSP = 1, 3, 5) in axial and coronal plane. The yellow dashed lines represent the border of the ROI for in vivo study. The residual magnetization inside the ROI is presented as mean ± standard deviation.

Validation of BESR sequence determined T1

The ¹H T1 measured in the tube phantom and in arterial blood flow is summarized as follows: BESR sequence determined ¹H T1 of circulating water (2.96 ± 0.43 s) was comparable to that of stationary water measured with a traditional saturation recovery sequence (3.30 ± 0.17 s, N.S.). The precision and accuracy errors were 13.0% and 10.3%, respectively. The precision and accuracy errors provide quantitative estimates for the potential measurement bias and uncertainty, respectively. The in vivo measured ¹H T1 in arterial blood (on 100% O₂) in left and right carotid arteries and left ventricle were 1.62 ± 0.12 s, 1.59 ± 0.2 s, and 1.67 ± 0.14 s, respectively. The corresponding in vitro T1 of sampled arterial blood saturated with 100% O₂ was 1.78 ± 0.18 s (N.S. compared to all in vivo measurements).

In vitro calibration of ¹⁹F R1 as a function of pO₂

In vitro measurements validated the linear dependence of ¹⁹F R1 on pO₂ (Fig. 3). At 37 °C,

$$R1(s^{-1})=(0.396\pm 0.014)+(0.0024\pm 0.0003)\times p{O}_2(\mathit{mmHg}).$$ [3]

Deoxy-Hb did not exhibit any observable paramagnetic relaxation enhancement effect on ¹⁹F T1, because ¹⁹F R1 vs pO₂ curve was identical in aqueous solution and mouse blood.

Figure 3.

Calibrated linear relationship between ¹⁹F R1 and pO₂ at room temperature and 37°. The calibration curve at 37° is used for the in vivo study. Error bars represent standard deviation.

In vivo measurement of blood oxygenation

Figure 4 shows representative BESR images of a mouse heart under hyperoxia and normoxia. After spin saturation, blood in the left ventricle (LV) exhibited a faster recovery of signal intensity than that in the right ventricle (RV), which was further accelerated when the mouse was breathing pure O₂ (i.e. hyperoxia). The fitted ¹⁹F R1 confirmed the faster relaxation of the more oxygenated blood. Based on the calibrated correlation between ¹⁹F R1 and pO₂ (Eq. 3), ¹⁹F MRI-measured LV blood pO₂ was 458 ± 133 mmHg under hyperoxia and 96 ± 21 mmHg under normoxia; significantly higher than RV blood pO₂, which was 103 ± 21 mmHg under hyperoxia and 49 ± 15 mmHg under normoxia (Fig. 5, p < 0.05 for all comparisons).

Figure 4.

(a) A short-axis ¹H image shows the left and right ventricles. (b) ¹⁹F BESR images of the same mouse heart under hyperoxia and normoxia. (c-d) Fitted ¹⁹F saturation recovery curve for LV and RV blood under normoxia and hyperoxia. Error bars represent standard deviation.

Figure 5.

Measured blood pO₂ based on measured ¹⁹F T1 of PFC NPs showed pO₂ difference between LV and RV, and between normoxic and hyperoxic conditions for each ventricle, *, p<0.05.

Accuracy of pO₂ measurements at different SNR

Since low SNR is a common problem in ¹⁹F MRI, the effect of SNR on ¹⁹F BESR determined pO₂ was evaluated by mathematical simulation (Fig. 6). As expected, the accuracy of ¹⁹F MRI measurements of pO₂ increased with SNR. However, the pO₂ measurement was less susceptible to noise (in the sense of absolute measurement uncertainty) at lower oxygen tension. When image SNR = 9, the measurement error of ¹⁹F MRI determined pO₂ was <10 mmHg under hypoxic condition (pO₂ <50 mmHg) and <40 mmHg under hyperoxic condition (pO₂ > 500 mmHg). Phantom studies with varying concentrations of PFC NPs (Fig. 6b) confirmed the simulation result. Higher imaging SNR substantially reduced the uncertainty and increased the precision of the ¹⁹F R1 measurements.

Figure 6.

(a) Simulation result shows measurement accuracy improves as SNR increases. pO₂ measurement uncertainty is higher under hyperoxic condition. (b) Measured pO₂ for three PFC NPs samples with different NP concentration bubbled with N₂, Air and O₂, respectively. The mean value of measured pO₂ is independent on PFC NPs concentration but measurement uncertainty (error bars) is suppressed by increased SNR.

Discussion

In the present work, we implemented a novel BESR sequence to quantify blood pO₂ based on ¹⁹F R1 of circulating PFC NPs. Because ¹⁹F R1 of PFC NPs is not affected by the paramagnetic effect of deoxy-Hb and manifests a linear association/release curve as a function of dissolved [O₂], it represents a promising probe to quantify blood pO₂. Our results demonstrate that the BESR sequence: 1) achieves a consistent T1 measurement for pulsatile blood flow, 2) differentiates blood pO₂ in the LV and RV, and 3) depicts the response of ventricular blood oxygenation to hyperoxic challenge.

One concern regarding the use of any oxygen sensing probe in vivo is whether other molecules or ions will affect the chemical properties of the probe. In this study, we observed that pO₂- ¹⁹F R1 relationship remained unchanged for PFC NPs in salt solution and blood sample. This is consistent with our previous findings that ¹⁹F R1 of PFC NPs is not affected by surrounding free paramagnetic ions (16,28) since their direct interactions are prevented by the lipid monolayer of PFC NPs, which is permeable to O₂ but not to macromolecules. The observed linear pO₂ response curve of PFC NP agreed with that observed in previous reports despite the difference in the fitted coefficient at different magnet field strength (12,29-30). The observed temperature dependence of ¹⁹F R1 has also been previously reported (15).

The long circulating half life time of PFC NPs and high ¹⁹F spin density are advantageous to use PFC NPs as a blood pool pO₂ probe. The mean half life of PFC NPs in the blood plasma is ~180 minutes in mouse and even longer in humans (31). Thus, the slow clearance of intravascular PFC NPs allows a relatively wide temporal window for quantitative pO₂ assessment with ¹⁹F MRI. In addition, even though PFC has high affinity for O₂, the modest administered dose is small enough such that the total amount of dissolved O₂ carried by PFC NPs is orders of magnitude less than that carried by plasma and red blood cells (dissolved plus oxy-hemoglobin) (32-34). Therefore, the blood-pool PFC NPs acted as a pO₂ sensor rather than a primary O₂ reservoir and the ¹⁹F MRI measured pO₂ values reflects pO₂ in blood stream.

In major arteries (or ventricles), and veins, the velocity of blood flow is > 10 cm/s and 1 cm/s, respectively. For a typical imaging slice thickness (1-2 mm), the minimal time required for blood to flow through the slice is approximately 100-200 ms. By setting the time interval between two consecutive excitation pulses greater than 200 ms and selecting imaging slices perpendicular to the flow direction, the previously excited spins should not affect the subsequent image acquisitions. Thus, gradient echo imaging with 90° excitation pulses can achieve maximal SNR without partial saturation effects. On the other hand, because the blood flow is rapid in major blood vessels, during the recovery process of the T1 measurement the real “imaged” spins within the region of interest may originate from distant locations. This requires homogeneous spin pre-saturation over a large volume of the imaged subject. Consistent with previous studies that utilized pre-saturation pulse trains (33,35), our result suggests that a non-selective saturation pulse train with crusher gradients was robust for reducing both B₀ and B₁ inhomogeneity while achieving a satisfactory uniform spin saturation profile. By implementing multiple saturation recovery acquisitions in each TR, the BESR sequence achieved higher SNR than traditional saturation recovery or inversion recovery T1 measurement sequences.

The novelty of BESR resides in the utilization of a homogenous RF pre-saturation and a train of ECG gated excitation/readouts in a single TR for fast T1 quantification. Due to the refreshing of intravascular PFC NP in the imaging slice (thickness = 2 mm) between adjacent heart beats, the acquired blood-pool ¹⁹F signal by each readout directly represents the ¹⁹F T1 recovery at corresponding delay times after saturation. In contrast, the conventional saturation pulse sequence only acquires one echo after each RF saturation, and the different T1 recovery weighting is achieved by varying TR (36). Thus, the BESR sequence provides a faster imaging method for quantifying blood ¹⁹F T1 by taking advantage of the time-of-flight effect of blood-flow in conjunction with intravascular PFC NP contrast agent. Although multiple 90° pulses were used in this study, other pulse shapes such as adiabatic or composite pulses also could be employed as long as homogeneous whole-body saturation is achieved (37). Similarly, fast imaging techniques like EPI could replace the single echo to further accelerate image acquisition (36).

The BESR sequence is optimized for ¹⁹F MRI because: (a) the PFC NP is an intravascular agent; (b) there is no ¹⁹F background in soft-tissue; (c) O₂ dissolving and release from PFC NP (~250 nm) operates through free diffusion that occurs in milliseconds (38). In capillaries, the measured pO₂ may directly reflect tissue pO₂ if the O₂ exchange between tissue and microvasculature is rapid (39). In ventricles and major blood vessel that experiencing no blood-tissue O₂ exchange, the measured pO₂ would represent venous or arterial blood pO₂. Finally, although the BESR sequence could be applied to quantify ¹H T1 in blood pool, the small correlation coefficient between ¹H R1 and pO₂ may be problematic for accurate pO₂ determination. Additionally, the O₂ release response curve of hemoglobin is non-linear, therefore the accuracy of ¹H MRI measured blood pO₂, as well as its relationship with tissue pO₂, remains to be defined.

The proposed ¹⁹F imaging methodology has direct translational potential. The technique might be readily adapted to evaluate arteriovenous pO₂ difference for selected organs, e.g. renal arteries and veins for kidney, carotid arteries and jugular veins for brain, aorta and coronary sinus for heart, among others. Further integration of ¹⁹F MRI determined A-V pO₂ with phase-contrast ¹H or ¹⁹F MRI determined blood flow velocity may enable assessment of oxygen consumption in individual organs as a critical pathological marker for evaluating injury at a cellular level.

Based on the accuracy profile of ¹⁹F-determined pO₂, imaging noise significantly affects the measurement uncertainty of pO₂ through non-linear T1 fitting. Therefore, the acquisition of ¹⁹F images with sufficient SNR is crucial to accurately estimate intravascular pO₂, which also was reported previously by Barker et al (14). In the present study, all images exhibited SNR > 7, and according to our simulation (Fig. 6a), pO₂ estimates from these images could achieve a measurement uncertainty as low as 10 mmHg under normoxia and 30 mmHg under hyperoxia depending on real pO₂.

Several limitations of the current pO₂ measurement paradigm should be mentioned. Even though the blood flow in major blood vessels is fast enough to create “time-of-flight” phenomenon, the velocity profile of blood flow is laminar, where the velocity close to the blood vessels walls is virtually zero. This portion of the blood may contribute to ¹⁹F signal resulting in reduced accuracy of pO₂ measurement in such regions (22). In the cardiac ventricles and major vessels, non laminar blood flow (e.g., turbulence) also may cause errors in T1 estimation due to spin de-correlation, which is not unexpected and can be visualized on ¹H images for proper data interpretation (40). Because the ejection fraction of normal ventricle is around 60%, after two heart beats around 16% residual blood still remains in the same ventricle. The residual blood may suffer from partial saturation from previous excitation and cause errors in T1 measurement. The systematic error (about 6% of T1 measured in phantom experiments) caused by these factors may be corrected by flow dynamical modeling in the future. The relative error of ¹⁹F BESR MRI determined blood pO₂, (i.e. 50 mmHg) in well oxygenated arterial blood or in venous blood (10 mmHg), could limit the application of this technique for measurement of tumor oxygenation, but there are other uses that offer insights into myriad pathophysiologies. Finally, for human imaging, where sampling once per cardiac cycle is too slow because of lower heart rate, the sequence should be modified to image more than once per cardiac cycle.