Quantitative ¹⁷O MRSI of myocardial oxygen metabolic rate, blood flow, and oxygen extraction fraction under normal and high workload conditions

Abstract

Purpose:

The heart is a highly aerobic organ consuming most of the oxygen the body in supporting heart function. Quantitative imaging of myocardial oxygen metabolism and perfusion is essential for studying cardiac physiopathology in vivo. Here, we report a new imaging method that can simultaneously assess myocardial oxygen metabolism and blood flow in the rat heart.

Methods:

This novel method is based on the ¹⁷O-MRSI combined with brief inhalation of ¹⁷O-isotope labeled oxygen gas for quantitative imaging of myocardial metabolic rate of oxygen consumption (MVO₂), myocardial blood flow (MBF), and oxygen extraction fraction (OEF). We demonstrate this imaging method under basal and high workload conditions in rat hearts at 9.4 T.

Results:

We show that this ¹⁷O MRSI–based approach can directly measure and image MVO₂ (1.35–4.06 μmol/g/min), MBF (0.49–1.38 mL/g/min), and OEF (0.33–0.44) in the heart of anesthetized rat under basal and high workload (21.6 × 10³–56.7 × 10³ mmHg • bpm) conditions. Under high workload condition, MVO₂ and MBF values in healthy rats approximately doubled, whereas OEF remained unchanged, indicating a strong coupling between myocardial oxygen metabolic demand and supply through blood perfusion.

Conclusion:

The ¹⁷O-MRSI method has been used to simultaneously image the myocardial metabolic rate of oxygen consumption, blood flow, and oxygen extraction fraction in small animal hearts, which are sensitive to the physiological changes induced by high workload. This approach could provide comprehensive measures that are critical for studying myocardial function in normal and diseased states and has a potential for translation.

1 |. INTRODUCTION

Cardiovascular disease is one of the leading causes of death worldwide. The heart is a highly aerobic organ that consumes large amounts of oxygen from the blood circulating in the body. Myocardial oxygen metabolism is tightly associated with the production of biochemical energy in the molecular form of adenosine triphosphate (ATP), which is critical to support the contraction/relaxation of cardiomyocytes to perform mechanical work under various working conditions. Many cardiovascular diseases have been linked to vascular abnormality and impaired energy metabolism. Therefore, it is important to exploit robust imaging approaches that can noninvasively and quantitatively measure myocardial metabolic rate of oxygen (MVO₂), ideally together with myocardial blood flow (MBF), then to determine another important physiology parameter: oxygen extraction fraction (OEF) in vivo.

With technological advancements over the past few decades, many biomedical imaging tools are available to noninvasively study different aspects of cardiac patho-physiology and function in animal models and humans (please see recent reviews on this topic^1,2 and the references cited therein). Of these, techniques based on MRS or MRSI and PET are the most relevant modalities for cardiometabolic imaging. However, for quantitative imaging of myocardial oxygen consumption rates in humans and/or large animals, ¹⁵O-PET is perhaps the only method that can directly measure and quantify MVO₂^3,4 and is considered the gold standard. Because the high-energy positrons emitted during ¹⁵O decay have a wide range, about 4 times that of the ¹⁸F isotope used in the popular fluorodeoxyglucose-18 (¹⁸FDG)–PET imaging method, the spatial resolution of ¹⁵O-PET is 4 times lower than that of ¹⁸FDG-PET.^5,6 Furthermore, the half-life of the ¹⁵O-isotope is only about 2 min,^5,6 so ¹⁵O-PET measurements rely on an on-site cyclotron, which is not available at most research or clinic sites.

On the other hand, the in vivo oxygen-17 (¹⁷O) MRS imaging (MRSI) approach has been successfully developed and demonstrated at ultrahigh magnetic field strength for 3D imaging and quantification of the cerebral metabolic rate of oxygen (CMRO₂) and possibly cerebral blood flow in rat and cat brains with a brief inhalation of ¹⁷O-isotope labeled oxygen gas (¹⁷O₂),^7–10 and for human brain applications.^11–14 In this study, we use 3D ¹⁷O MRSI to evaluate the feasibility of quantitatively imaging regional MVO₂ as well as MBF and OEF in the rat heart based on a single imaging measurement with short (2 min) inhalation of ¹⁷O₂ at 9.4 T under both basal and high workload conditions.

2 |. METHODS

2.1 |. Principle and methodology

The principle of ¹⁷O-MRSI to measure oxygen metabolism is similar to the 15O-PET imaging method: Both approaches involve the inhalation of isotope-labeled oxygen gas (¹⁷O₂ for 17O-MRSI and 15O₂ for 15O-PET imaging), followed by the determination of the rate at which inhaled ¹⁷O-labeled or ¹⁵O-isotope-labeled oxygen is converted into ¹⁷O-labeled or ¹⁵O-labeled water in tissues through oxygen metabolism, which is the oxygen metabolism rate.^8,9,13,15 However, the ¹⁷O-MRSI method has several advantages over the ¹⁵O-PET method. The natural abundance of ¹⁷O in oxygen gas is very low (0.037%) but can be enriched to >70%; this highly enriched oxygen gas is non-radioactive, stable, and commercially available. Moreover, ¹⁷O₂ in tissue is “invisible” to in vivo ¹⁷O-MRS/MRSI detection, and ¹⁷O-MRS/MRSI can only detect the final oxygen metabolism product—¹⁷O-labeled water H ¹⁷₂O—thus, making imaging measurement and quantification simple.^8,9 Metabolism of normal (or natural-abundance) oxygen in inspired air, dominated by ¹⁶O₂, does not alter the H ¹⁷₂O signal in ¹⁷O MRS/MRSI.

Figure 1A illustrates the setup for an animal study in which a simple device was used to allow rapid switching between an inspired gas mixture containing normal natural-abundance oxygen gas and a gas mixture containing ¹⁷O-enriched oxygen gas, which was delivered to the animal’s lungs through a ventilator (note: a ventilator may not be required for spontaneously breathing animals). For an aerobic organ with high oxygen metabolic rate and blood perfusion (such as heart or brain), the detected ¹⁷O-MRSI signal and time-dependent changes in tissue H₂¹⁷O can be characterized by three phases, as shown in Figure 1B:

• Phase I (before inhalation): Natural-abundance ¹⁷O-MRS signal of tissue H ¹⁷₂O; this can be used as an internal reference for quantifying tissue H₂¹⁷O concentration and its changes in the subsequent two phase periods.

• Phase II (¹⁷O₂ inhalation): ¹⁷O-MRS signal during a brief ¹⁷O₂ inhalation (usually lasting a few minutes), showing a rapid increase of tissue H ¹⁷₂O content through the metabolism of inhaled ¹⁷O₂ in cells; the dynamic H₂¹⁷O change can be used to quantify MVO₂.

• Phase III (after inhalation): 17O-MRS signal after the ¹⁷O₂ inhalation, in which H₂¹⁷O molecules produced via the metabolism of inhaled ¹⁷O₂ in tissue begin to decay due to washout by blood perfusion; they can be used as an endogenous perfusion tracer to measure and quantify MBF.¹⁶

FIGURE 1.

(A) Schematic of a simple setup that allows rapid switching between a natural-abundance oxygen gas mixture and an ¹⁷O-enriched oxygen gas mixture sent to the animal’s lung by a ventilator and used to perform cardiac ¹⁷O-MRS imaging in rats. (B) For organs with high oxygen metabolic rate and blood perfusion, such as brain or heart, the dynamics of cardiac H₂¹⁷O signal can be characterized by three phases. Phase I: Steady-state natural-abundance ¹⁷O-water signal before ¹⁷O₂ inhalation. Phase II: A brief period of ¹⁷O₂ inhalation (indicated by a gray bar, usually a few minutes), resulting in a rapid increase in tissue H₂¹⁷O signal. Phase III: Following the ¹⁷O₂ inhalation period, the metabolized H₂¹⁷O in myocardial tissue begins to decay and can serve as an endogenous perfusion tracer.

The mass balance relationship for the concentration of labeled H₂¹⁷O in myocardium during inhalation of ¹⁷O₂ gas is given as follows⁹:

$$\frac{d{C}_{\text{m}}(t)}{dt}=2\cdot \alpha \cdot f_1\cdot {\text{MVO}}_2+f_2\cdot \text{MBF}\left[C_{\text{a}}(t)-C_{\text{v}}(t)\right]$$ (1)

where C_a, C_m, and C_v are the H₂¹⁷O concentration expressed in excess of the natural abundance H₂¹⁷O concentration in arterial blood, myocardial tissue and venous blood, respectively, as a function of the ¹⁷O₂ inhalation time t (unit = minute). One labeled ¹⁷O₂ converts to two H₂¹⁷O molecules through oxidative metabolism; thus, a constant of 2 is used in Eq. (1). The parameter α in Eq. (1) is the ¹⁷O enrichment fraction in the inhaled ¹⁷O₂ gas. Natural-abundance H₂¹⁷O concentration (equal to 20.35 μmol per gram tissue water or per gram water in blood [see Ref. 9]) can be used to calibrate the absolute concentrations of C_m(t), C_a(t), and C_v(t). For convenience, the unit for C_m is chosen as μmol/(g heart tissue water), and the units for C_a and C_v are μmol/(g blood water), where parameters f₁ (=1.266) and f₂ (=1.077) are conversion factors used for consistency of units applied in Eq. (1). Because the tissue water composition is similar between brain and heart,¹⁷ the same f₁ and f₂ conversion factors can be applied to brain and heart tissues.⁹

2.2 |. Quantification of MVO₂ (Phase II)

According to the Taylor’s Theorem, C_m(t) during ¹⁷O₂ inhalation can be written as

$$C_{\text{m}}(t)=a_0+a_1\cdot t+a_2\cdot t^2+a_3\cdot t^3+\dots .$$ (2)

Because C_m is zero at the beginning of the ¹⁷O₂ inhalation (i.e., C_m[0] = 0); thus, a₀ = 0. Similarly, the second term on the right side of Eq. (1) can also be written as

$$f_2\cdot \text{MBF}\left[C_a(t)-C_v(t)\right]=b_0+b_1\cdot t+b_2\cdot t^2+b_3\cdot t^3+\dots$$ (3)

with the boundary condition that both C_v and C_a are zero at the start of the inhalation (i.e., C_v[0] = C_a[0] = 0). Therefore, Eq. (1) becomes

$$a_1+2\cdot a_2\cdot t+3\cdot a_3\cdot t^2+\dots =2\cdot \alpha \cdot f_1\cdot {\text{MVO}}_2+b_1\cdot t+b_2\cdot t^2+b_3\cdot t^3+\dots$$ (4)

Because Eq. (4) holds for any t value, applying the Polynomial Theorem to Eq. (4) at t = 0 leads to

$$a_1\approx 2\cdot \alpha \cdot f_1\cdot {\text{MVO}}_2$$ (5)

Thus,

$${\text{MVO}}_2\approx \frac{a_1}{2\cdot \alpha \cdot f_1}.$$ (6)

When C_m(t) is regressed by a polynomial function in the form of Eq. (2) with a₀ = 0, the fitting coefficient of the linear term (i.e., a₁) can be used to calculate MVO₂ according to Eq. (6).

In principle, the higher the order of the polynomial used, the better the fit. However, when fitting a polynomial equation on experimental data with a limited SNR, the reliability of the fitting results is poor due to the overfitting of the data to polynomials of larger degrees; therefore, in practice, only linear and quadratic terms in Eq. (2) were used to fit the C_m(t) time courses for each imaging voxel to improve the fitting reliability. Our previous work showed that quadratic fitting provides accurate results for estimating metabolic rate of oxygen in rat brains, in contrast to linear fitting that could underestimate metabolic rates by up to 10% for rat brain data with several minutes of ¹⁷O₂ inhalation.¹⁸ In this study, we use a quantification approach based on the quadratic fitting.

2.3 |. Quantification of MBF (Phase III)

During Phase III (after inhalation), the decay of the tissue H₂¹⁷O signal or concentration approximately follows the following exponential function:

$$C_m(t)=C_1\times e^{-kt}+C_2$$ (7)

Based on our early work of rat brains under various conditions,¹⁶ the determined decay rate (k) is approximately 1.86 times slower than that measured by the classic tracer method using a bolus injection of H₂¹⁷O into a carotid artery; thus,

$$\text{MBF}(\text{mL}/\text{g}/\text{min})\approx 1.86(\text{mL}/\text{g})\times \text{k}\left({\text{min}}^{-1}\right)$$ (8)

2.4 |. Quantification of OEF

$$\text{OEF}=\frac{{\text{MVO}}_2}{{\text{C}}_{\text{a},{\text{O}}_2}\times \text{MBF}}\times 2.24$$ (9)

where C_a,O2 is the artery oxygen concentration and is treated as a constant. The common unit of the C_a,O2 is mL-O₂/dL-blood and can be converted to μmol/mL-blood for matching the unit of MVO₂ to MBF ratio according to Avogadro’s law, and the conversion factor is 2.24. In this study, we used the C_a,O2 value of 15.1 mL-O₂/dL-blood measured in healthy and young rats as reported in the literature¹⁹ to calculate the myocardial OEF in the rat heart.

2.5 |. Materials and experiment

Eight Sprague–Dawley rats (body weight: 318 ± 38 g) were used in this study. The rats were placed in an induction box under 5% isoflurane for several minutes, after which the animals were intubated and mechanically ventilated under gaseous anesthesia (2% isoflurane in a mixture of 70% nitrous oxide and 30% oxygen). The femoral artery and vein were cannulated for physiological monitoring, blood sampling, and chemical administration. The animals’ rectal temperature was maintained at 37.1 ± 0.3°C using heated circulating water; the arterial blood pressure and heart rate were monitored and maintained under normal physiological conditions throughout the experiment. The ventilated oxygen gas could be switched between natural-abundance (most time) and ¹⁷O-enriched (up to 89% ¹⁷O enrichment and for 2 min of inhalation) O₂ gas using the device shown in Figure 1A. The animals were studied under either basal and/or high workload condition using an intravenous bolus injection of 40 μg/(kg body weight) dobutamine and dopamine.

To achieve high SNR and temporal resolution for in vivo ¹⁷O-MRS/MRSI data in the rat heart, we developed a novel RF coil configuration consisting of two electrically decoupled RF coils. A small ¹⁷O surface coil (~1 cm diameter) with electric isolation and wrapped with soft Teflon film around the coil plane was inserted into the rat chest via thoracotomy and placed between the chest (inner) wall and the heart for holding the heart organ as illustrated in Figure 2A; the incision of the muscles was closed by suture after implanting the coil. Another half-volume proton (¹H) RF coil was positioned around the rat’s body while the animal was in prone position with a RF magnetic field (B₁) perpendicular to that generated by the ¹⁷O surface coil, resulting in passive decoupling between the two RF coils. The inserted ¹⁷O surface coil has two advantages: It provides a higher SNR for the detection of the signals from the heart due to the proximity between the RF coil and imaging object, enabling higher spatiotemporal imaging resolution; it also significantly reduces the contamination signals from surrounding chest muscle and the heart motion induced by breathing and lung motion. No additional gating or motion control was applied in this study. The University of Minnesota Institutional Animal Care and Use Committee approved the animal protocols used in this study.

FIGURE 2.

(A) Representative ¹H MRI of rat chest and heart tissues in coronal (left panel) and sagittal (right panel) orientations, the location of the ¹⁷O surface coil is depicted on the images. (B) Stack plots of global heart H₂¹⁷O resonance signals (3 s per ¹⁷O spectrum) and the corresponding H₂¹⁷O signals time course measured before, during, and after a 2-min inhalation (indicated by gray bars) of ¹⁷O₂ gas. D, dorsal; F, foot; H, head; L, left; R, right; V, ventral.

All ¹⁷O-MRS/MRSI experiments were conducted on a 9.4T/31-cm bore animal magnet (Magnex Scientific, Oxford UK) with a Varian INOVA console (Varian Inc., Palo Alto, CA, USA). The ¹H RF half-volume coil was operated at 400 MHz, and the ¹⁷O RF surface coil was operated at 54.3 MHz.

A 3D-CSI method based on the Fourier series window (FSW) k-space weighting^20,21 was applied in this study. The FSW weighting applied for the CSI acquisition is based on the Fourier coefficients of a predetermined voxel shape following the optimal termination of the series. The FSW-based CSI approach is characterized by better sensitivity and smaller out-of-voxel contamination as compared with the conventional Fourier transform approach.^20,21 The 3D-CSI acquisition parameters were 9 × 9 × 5 phase encodes, 3 × 3 × 2.5 cm³ FOV, 3.3 × 3.3 × 5 mm³ nominal voxel size, 45-ms TR, 30-kHz spectral width, 11-s acquisition time per 3D ¹⁷O CSI. Multiple CSI volumes were collected to capture the dynamic H₂¹⁷O signal change in the heart across the three phases as illustrated in Figure 1B. A hard RF pulse was applied to excite the ¹⁷O water resonance, and its power was calibrated using a single-pulse acquisition sequence with varied RF pulse power to optimize the ¹⁷O signal of the natural abundance water dominated by the rat heart. The FID from each CSI voxel was zero-filled before performing Fourier transformation, and a line broadening (100 Hz) was applied to enhance SNR.

3 |. RESULTS

Figure 2A shows ¹H structural images in two orientations covering the rat chest, heart, and other adjacent tissues. The position of the ¹⁷O surface coil (orange in Figure 2A) is marked schematically in the image, indicating its close proximity to the heart, and is used to detect the ¹⁷O water signal primarily from the myocardium. Figure 2B displays a stack plot of global myocardial ¹⁷O water (H₂¹⁷O) spectra with a high temporal resolution of 3 s per ¹⁷O spectrum and superior sensitivity, as well as the corresponding H₂¹⁷O signal time course before, during, and after a 2-min ¹⁷O₂ gas inhalation. The three distinct phases of myocardial H₂¹⁷O signal changes are highly similar to those observed in the rodent brains.^7,9,18,22,23 The first phase (before inhalation) measures the natural-abundance ¹⁷O-water signal from the myocardium, which can be used to calibrate the absolute H₂¹⁷O concentration^7,9,18,22,23 in the myocardium for quantifying MVO₂. The second phase is characterized by rapid accumulation of myocardial H₂¹⁷O content, primarily ¹⁷O-labeled water produced in mitochondria by myocardial ¹⁷O₂ metabolism; the accumulation rate is tightly correlated to MVO₂. The third phase shows that after ¹⁷O₂ inhalation is terminated, the H₂¹⁷O signal decays as labeled H₂¹⁷O water is washed away from the myocardium via perfusion; its decay rate constant (k = 0.56 min⁻¹ in this measurement) is related to myocardial blood perfusion (i.e., MBF). According to (Eqs. 6) and (8), for this particular measurement, the values of global heart MVO₂ and MBF were determined to be 1.9 μmol/g/min and 1.1 mL/g/min, respectively.

Figure 3A presents the imaging results from another rat, showing the H₂¹⁷O time courses with 2 min of ¹⁷O₂ inhalation from four representative ¹⁷O-CSI voxels located inside the heart and the corresponding 1H rat heart image. Excellent sensitivity and high temporal resolution (11 s per CSI volume) of the ¹⁷O-CSI make it possible to image the dynamics of the myocardial H₂¹⁷O in situ. The H₂¹⁷O time courses of these 4 voxels were well characterized with three distinct phases and relatively small signal fluctuation, resulting in MVO₂ values in the range of 1.82–2.03 μmol/g/min and MBF values in the range of 0.47–0.65 mL/g/min, respectively. Figure 3B shows the ¹⁷O water signal maps of the same ¹⁷O-CSI slice at three time points imaged before/after the ¹⁷O₂ inhalation (see Figure 3A for time labels), demonstrating the spatial distribution of the H₂¹⁷O signal in the ¹⁷O-RF coil sensitive region and its changes in response to ¹⁷O₂ inhalation, metabolism and washout.

FIGURE 3.

(A) Cardiac ¹H MRI of a representative rat with heart outlined by a yellow dashed line; the locations of four selected CSI voxels inside the heart from a representative CSI slice are marked by dashed circles (center panel). The H₂¹⁷O signal time courses of these voxels and the corresponding metabolic rate of oxygen consumption (MVO₂) and myocardial blood flow (MBF) values are displayed for each selected voxel. (B) H₂¹⁷O signal maps of the same CSI slice at different time points before and after ¹⁷O₂ gas inhalation, showing the distribution of H₂¹⁷O signals in the cardiac region covered by the ¹⁷O RF surface coil, and the change of H₂¹⁷O signal before and after ¹⁷O₂ inhalation. Time labels and inhalation period are shown in (A, left-top panel). F, foot; H, head; L, left; R, right.

Figure 4 illustrates the results based on the average of multivoxel data from three ¹⁷O-CSI slices and the values of MVO₂ and MBF in each slice, indicating a similar trend of these changes along the slice direction (averaged values from the three slices: MVO₂ = 1.88 μmol/g/min and MBF = 0.60 mL/g/min).

FIGURE 4.

Cardiac ¹H MRI in two orientations, showing the location of the ¹⁷O-RF surface coil and the positions of three representative ¹⁷O-CSI slices and the metabolic rate of oxygen consumption (MVO₂) and myocardial blood flow (MBF) values from the averaged multivoxels for each ¹⁷O-CSI slice. D, dorsal; F, foot; H, head; L, left; R, right; V, ventral.

We also measured the dynamics of the myocardial H₂¹⁷O signals and calculated the corresponding MVO₂ and MBF values in the same rat heart under basal and high workload conditions. Two examples are presented herein. Figure 5 shows the results based on the global H₂¹⁷O signals (only used ¹⁷O surface coil localization), and Figure 6 displays the time courses of a representative CSI voxel located inside the heart and the H₂¹⁷O maps of a selected CSI slice across the rat heart at different times and conditions. Apparently, infusion of mixed dobutamine and dopamine largely increased the blood pressure and heart rate, leading to a high workload condition with greatly increased heart-rate pressure product (RPP, in mmgH·bpm bpm). Accordingly, the myocardial ¹⁷O water signal increased faster during the ¹⁷O₂ inhalation and reached a higher peak at the end of the ¹⁷O₂ inhalation (Phase II as illustrated in Figure 1B) under the high workload condition, resulting in a higher MVO₂ rate (e.g., MVO₂, high-workload = 3.18 μmol/g/min vs. MVO₂, basal = 1.78 μmol/g/min as shown in Figure 5B,C). The H₂¹⁷O signal also decayed faster during the post-inhalation period (Phase III as illustrated in Figure 1B) under the high workload condition, indicating a higher perfusion rate (e.g., MBFhigh-workload = 1.37 mL/g/min vs. MBFbasal = 0.71 mL/g/min as shown in Figure 5B,C). Similar increases of MVO₂ and MBF were also detected in a representative 3D 17O-CSI voxel located inside another rat heart (cycled in Figure 6A), showing again significantly elevated MVO₂ = 2.66 μmol/g/min and MBF = 1.12 mL/g/min under the high workload condition as compared to the basal condition with MVO₂ = 1.37 μmol/g/min and MBF = 0.64 mL/g/min. These results collectively show that myocardial oxygen consumption rate and perfusion approximately double as myocardial workload increases and suggest a tight coupling relationship between myocardial oxygen metabolism and perfusion under basal and high workload conditions.

FIGURE 5.

(A) Illustration of representative ¹H cardiac MRI (in axial and sagittal orientations) of another rat and the location of the ¹⁷O-RF surface coil. Stack plots of global heart H₂¹⁷O resonance signals (3 s per spectrum) and the corresponding H₂¹⁷O time courses measured before, during, and after a 2-min inhalation (indicated by gray bars) of ¹⁷O₂ gas under basal (B) and high workload (C) conditions. The values of metabolic rate of oxygen consumption (MVO₂), myocardial blood flow (MBF), and rate pressure product for each condition are displayed. D, dorsal; F, foot; H, head; L, left; R, right; V, ventral.

FIGURE 6.

(A) Cardiac ¹H MRI and the myocardial H₂¹⁷O time course of a representative 3D ¹⁷O-CSI voxel inside the rat heart measured before, during, and after a brief (2-min) ¹⁷O₂ inhalation under basal and high workload conditions. The position of heart and voxel are indicated by a dashed outline and circle, respectively, on the ¹H MRI; the values of metabolic rate of oxygen consumption (MVO₂), myocardial blood flow (MBF), and rate pressure product (RPP) for each condition are also displayed. (B) H₂¹⁷O signal maps of the same CSI slice at different time points before and after ¹⁷O₂ gas inhalation under basal and high workload conditions. Time labels and inhalation period are shown in (A). F, foot; H, head; L, left; R, right.

With the MVO₂ and MBF values of the same heart under the same condition, we can readily calculate OEF of the myocardial tissue following (Eq. 9). The individual and averaged MVO₂, MBF, OEF, and RPP of all eight rats studied are summarized in Table 1. We found statistically significant increases in mean MVO₂ and MBF under high workload conditions compared with basal conditions. Figure 7 shows the correlation analysis results of MVO₂ versus MBF (Figure 7A), MVO₂ versus RPP (Figure 7B), MVO₂ versus OEF (Figure 7C), as well as their linear regression curves, resulting in the following two approximated equations:

$$\text{RPP}(\text{mmgH}\cdot \text{bpm})\approx 1.40\times {10}^4\times {\text{MVO}}_2\left(R^2=0.37\right)$$ (10)

$$\text{MBF}(\text{mL}/\text{g}/\text{min})\approx 0.39\times {\text{MVO}}_2\left(R^2=0.93\right)$$ (11)

TABLE 1.

Summary of the measurement results in healthy rats (n = 8).

	Basal condition				Higher workload
Rats	MVO2 (μmol/g/min)	MBF (mL/g/min)	OEF	RPP $(\text{mmHg}\cdot \text{bpm})$	MVO2 (μmol/g/min)	MBF (mL/g/min)	OEF	RPP $(\text{mmHg}\cdot \text{bpm})$
1	1.83	0.81	0.33	36 586	2.41	0.86	0.42	47 239
2	1.71	0.65	0.39	42 211	-	-	-	-
3	1.35	0.49	0.41	27 530	-	-	-	-
4	2.80	1.24	0.33	24 609	4.06	1.38	0.44	36 000
5	2.80	0.96	0.43	35 711	2.88	1.22	0.35	41 622
6	1.89	0.71	0.39	21 677	1.97	0.73	0.40	25 046
7	1.78	0.71	0.37	25 818	3.18	1.36	0.35	56 680
8	2.08	0.83	0.37	26 867	2.24	0.91	0.37	34 656
Mean	2.03	0.80	0.38	30 126	2.79 *	1.08 *	0.39	40 207 *
SD	0.52	0.23	0.03	7138	0.76	0.28	0.04	10 965

^*Indicating statistically significant changes with p < 0.05 (Student t-test) were detected between the two conditions.

Abbreviations: MBF, myocardial blood flow; MVO₂, metabolic rate of oxygen consumption; OEF, oxygen extraction fraction; RPP, rate pressure product.

Note: Only the listed values with the * sign indicates statistically significant for the high working load condition as compared to the basal condition with a p value of < 0.05 (in bold).

FIGURE 7.

Regression analysis results showing a strong linear correlation between cerebral blood flow (CBF) and metabolic rate of oxygen consumption (MVO₂) (A), a positive correlation between rate pressure product (RPP) and MVO₂ (B), and oxygen extraction fraction (OEF) values were insensitive to changes in MVO₂ in all rats studied (C). MBF; myocardial blood flow.

We also found that the mean OEF values (OEFbasal = 0.38 ± 0.03 and OEF_{high-workload} = 0.39 ± 0.04) are almost identical under basal and high workload conditions with RPP_basal = (3.0 ± 0.7) × 10⁴ mmHg·bpm and RPP_{high-workload} = (4.0 ± 1.1) × 10⁴ mmHg·bpm, respectively, in the rat hearts anesthetized using isoflurane. On average, a 33% increase in RPP (from 3.0 to 4.0 × 10⁴ mmHg·bpm) resulted in a 37% increase in MVO₂ (from 2.03 to 2.79 μmol/g/min) and 35% increase in MBF (from 0.80 to 1.08 mL/g/min) (see details in Table 1). Together, these results demonstrate that the ¹⁷O-MRSI method is sensitive and robust for quantitatively imaging changes in myocardial metabolism and perfusion in vivo in a noninvasive manner.

4 |. DISCUSSION

Like the brain, the heart relies on extensive ATP energy production from oxidative metabolism to support normal cardiac function. Most of the ATP molecules used by the normal adult heart are produced by the oxidation of fatty acids in the mitochondria, and the remainder is by the oxidation of glucose.^24–26 Because the P:O (phosphorylation vs. oxygen metabolism) ratio of fatty acid oxidation is relatively lower than that of glucose oxidation,²⁷ for the same ATP synthesis rate, more oxygen must be consumed when fatty acids are used as substrates for energy metabolism. A healthy human heart can synthesize approximately 6 kg of ATP molecules per day,²⁸ which is similar to the 5.7 kg of ATP molecules produced by the human brain per day,²⁹ although the heart weight is 3–4 times lighter than the brain. The high activities of oxygen metabolism and substrate oxidation in the heart are closely associated with high coronary blood supply and tissue perfusion to meet the high oxygen demand of the heart even in the basal state.³⁰ Blood flow and oxygen consumption are highly dynamic and tightly coupled under working state, as shown in Figure 7A. Abnormality in energy metabolism is associated with cardiac dysfunction and diseases; for example, obesity and diabetes have increased myocardial oxygen consumption and fatty acid oxidation rates, resulting in reduced cardiac efficiency.³¹ Therefore, quantitative and robust imaging of myocardial oxygen metabolism, blood flow, and oxygen extraction fraction is essential for assessing and understanding cardiac physiopathology.

The ¹⁵O-PET imaging method was established decades ago to image oxygen consumption and blood flow rates and OEF in the brain³² or heart.^3,4 By inhaling oxygen gas enriched with radioactive ¹⁵O-isotope (¹⁵O₂) and monitoring the spatial distribution and accumulation rate of ¹⁵O-isotope labeled signals in tissues, ¹⁵O-PET can determine the regional metabolic rate of oxygen consumption. However, ¹⁵O-PET is unable to distinguish the radioactive signals between the ¹⁵O₂ molecules bound to hemoglobin and the ¹⁵O atoms incorporated into H₂¹⁵O. To overcome this shortcoming, two additional measurements are commonly required to determine blood volume with inhaled C¹⁵O and blood flow with an intravenous H₂¹⁵O tracer.³² Therefore, three ¹⁵O-PET scans with different radioactive tracers are required to quantitatively determine the metabolic rate of oxygen consumption, blood flow, and OEF based on a complex model with a number of approximations and assumptions. This presents challenges for longitudinal studies with multiple measurements due to the safety concern of excess radiotracers taken by the body.

In this pilot study, we demonstrate for the first time the feasibility of a novel 3D ¹⁷O-MRSI method for quantitative imaging of three important quantities: MVO₂, MBF, and OEF via a single imaging measurement with a brief (2 min) inhalation of ¹⁷O-labeled oxygen gas (¹⁷O₂). Although ¹⁵O-PET and ¹⁷O-MRSI imaging methods are similar in methodology and principle, ¹⁷O MRSI offers merits. First, the ¹⁷O isotope is stable and non-radioactive. Second, ¹⁷O MRSI specifically detects the metabolically produced H₂¹⁷O from inhaled ¹⁷O₂ without the confounding signals from ¹⁷O₂ molecules bound to hemoglobin or dissolved in tissues, so the ¹⁷O signal is only detectable when ¹⁷O₂ is incorporated into water via oxygen metabolism.^8,9 Because of this unique specificity, quantification of MVO₂ from ¹⁷O-MRSI data becomes simpler (see Section 2) compared with the ¹⁵O-PET imaging method. Third, the metabolically generated H₂¹⁷O in the myocardium can serve as an endogenous perfusion tracer for determining MBF. These merits allow the simultaneous imaging of MVO₂, MBF, and OEF in small animals with a single noninvasive ¹⁷O-MRSI measurement and a brief ¹⁷O₂ inhalation. Moreover, the sensitivity of ¹⁷O MRSI is significantly improved using high or ultrahigh-field MRI scanner,^33,34 enabling high spatiotemporal imaging resolution and high-quality images, as illustrated in this study at 9.4 T.

To reduce the heart motion caused by breathing, and achieve optimal detection sensitivity, we designed an implantable ¹⁷O-RF surface coil to hold the rat heart when performing the dynamic ¹⁷O-MRSI study with high spatiotemporal resolution. From the structural ¹H MRI and 3D ¹⁷O MRSI (Figures 3 and 6), the global signal detected by the ¹⁷O surface coil was dominated by the heart with negligible contamination from surrounding chest muscle. Therefore, the derived parameters based on the global H₂¹⁷O signal and its dynamic change should primarily reflect the metabolic rate of oxygen consumption, blood flow, and OEF of the heart (Figures 2 and 5). Although global H₂¹⁷O signal does not provide regional information within the heart, it offers excellent sensitivity to monitor the dynamic change of the myocardial tissue H₂¹⁷O metabolized from inhaled ⁷O₂ (Figures 2B and 5B,C), thereby robustly determining the MVO₂ and MBF values.

The ¹⁷O-RF coil configuration and high field strength at 9.4 T provide excellent detection sensitivity, making it possible to perform 3D ¹⁷O MRSI with relatively high spatial (nominal voxel size: 0.054 mL) and temporal (11 s per 3D ¹⁷O-MRSI volume) resolution (Figures 3 and 6). Considering that the heart weight of the Sprague–Dawley rats is about 1.1 g,³⁵ a myocardial tissue density is 1.1 g/mL, the myocardial tissue volume is estimated to be about 1.0 mL, and thus covered by up to 18 ¹⁷O-CSI voxels. Nevertheless, we found that the MVO₂ and MBF values from four representative voxels as shown in Figure 3A (MVO₂: 1.82–2.03 μmol/g/min; MBF: 0.47–0.65 mL/g/min) or the averaged values from three 17O-CSI slices as shown in Figure 4 (MVO₂ = 1.88 μmol/g/min; MBF = 0.65 mL/g/min) are in an excellent agreement with the global measurement results from the same animal (MVO₂ = 1.89 μmol/g/min; MBF = 0.71 mL/g/min, see Rat 6 in Table 1). The lack of regional differences in the measured MVO₂ and MBF values may be due to partial volume effects (see more discussion later).

The ¹⁷O-MRSI measurement is sensitive to the heart workload. This study shows approximately doubling the MVO₂ and MBF values under the high workload condition; for instance, MVO₂ and MBF increased 79% and 93%, respectively, based on the global measurement in an isoflurane anesthetized rat (Figure 5), and increased 94% and 75%, respectively, in a representative voxel in the heart of a different rat (Figure 6). The results are consistent with an isolated and perfused rat heart study, showing a 92% increase of MVO₂ at high workload condition using high perfused Ca²⁺ concentration.³⁶ The coupling relationship between MVO₂ and the workload strength as quantified by the RPP parameter is shown in Figure 7B, and a similar correlation was also reported in literature.^3,4 As expected in healthy hearts, there was a strong, linear correlation between MVO₂ and MBF measured under basal and high workload conditions (Figure 7A and Eq. [11]), resulting in OEF values (OEFbasal = 0.38 ± 0.03 and OEFhigh-workload = 0.39 ± 0.04) that are insensitive to changes in MVO2 and MBF under normal physiology conditions.

It has been reported in the literature that the MVO₂ values of normal human^3,37,38 and anesthetized dog^4,39–41 hearts measured by conventional methods are about 4.5–6.5 and 3.2–4.7 μmol/g/min, respectively; however, due to the difficulty of measuring MVO₂ in small animals, there is a lack of MVO₂ values for normal rats in the literature. The MVO₂ (1.35–4.06 μmol/g/min) obtained in healthy rats in this study is similar to the range of humans and dogs, but lower than expected given the higher metabolic activity of rodents. Similarly, the MBF (0.49–1.38 mL/g/min) and OEF (0.33–0.44) obtained in this work are also lower than those reported values of healthy humans with MBF_human ≈ 0.4–1.2 mL/g/min and OEF_human ≈ 0.6–0.7,^{3,37,38,42,43} dogs with MBF_dog ≈ 0.6–1.3 mL/g/min and OEF_dog ≈ 0.5–0.6^4,40,41,44 or rats with MBF ≈ 1.5–6.5 mL/g/min.^45–48 rat The main reason for this underestimation in the present study is the uncorrected partial volume effect of the intracardiac blood pool, although in cardiac PET studies, the systematic underestimation in measurements of myocardial radiotracer concentration due to wall motion and small transmural wall thickness has been corrected.^49,50 If one assumes a myocardial tissue fraction of 0.5 in the region of interest or imaging voxel, and that the dynamic change in H₂¹⁷O concentration in the blood pool is half that of the myocardium during the ¹⁷O₂ inhalation, a 3-fold increase in MVO₂, 50% or smaller increase in MBF and about 2-fold increase in OEF can be anticipated. It seems more reasonable if a similar correction is applied to the results obtained in the current study. Further work is needed to improve quantification.

In the present study, we independently processed Phases II and III of the myocardial H₂¹⁷O time course to quantify MVO₂ and MBF, respectively (see Section 2). However, for organs or tissues with low oxygen metabolic rate and blood flow, such as skeletal muscle at rest, or even human brain or heart, a sophisticated quantification model combining Phases II and III (i.e., a two-phase model) is needed to correctly determine myocardial oxygen consumption rate^13,23 or to measure CMRO₂, cerebral blood flow, and thus OEF in the human brain.¹⁴ Therefore, for human cardiac imaging application using the ¹⁷O-MRSI method, a two-phase quantification model should be considered.

We designed and applied an implantable ¹⁷O-RF surface coil on a rat model to achieve an excellent SNR for 3D ¹⁷O-MRSI measurements; however, a small surgery for coil implantation is required. Surgery can be avoided if the ¹⁷O-RF coil is placed outside the chest, which is preferable for close-chest or human applications, although SNR may be reduced. Recently, we demonstrated the feasibility of performing a completely noninvasive deuterium (²H) MRSI (DMRSI) study in a close-chest rat heart model⁵¹; and a novel machine learning–based method that greatly improves SNR and enables high-resolution dynamic DMRSI study of brain tumor in rats.⁵² We believe the potential of the ¹⁷O-MRSI technique for non-invasive imaging MVO₂, MBF and OEF in the human heart in normal and/or diseased states can be realized if these two approaches are combined with a 3D ¹⁷O-MRSI method. The ¹⁷O MRSI or ¹⁷O MRI has been applied to map human brain CMRO₂ at varied magnetic field strength from 3 T to 9.4 T.^11–14,53 It has also been shown that the natural-abundance H₂¹⁷O signal (i.e., Phase I) in the human myocardium can be imaged at the clinical MRI strength of 3 T.⁵³ Therefore, by introducing a biphasic quantification model¹⁴ and a machine learning–based denoising method in a ¹⁷O-MRSI approach capable of simultaneous MVO₂, MBF and OEF imaging, one should be able to apply this technique to human cardiac imaging.

McCommis et al. has reported an alternative myocardial imaging approach based on the proton T_1ρ MRI for indirectly detecting the change of the myocardial H₂¹⁷O content metabolized from the injected ¹⁷O₂-enriched artificial blood, and thus to determine MVO₂ in canine.⁵⁴ The indirect ¹⁷O method based on proton imaging read-out is expected to provide higher detection sensitivity than the direct ¹⁷O imaging method; however, in practice the reported method has several technical limitations and drawbacks. First, it is difficult to accurately quantify tissue H₂¹⁷O concentration and MVO₂ value due to low sensitivity in detecting proton signal changes that are directly related to changes in tissue H₂¹⁷O content, and it requires multiple T_1ρ MRI measurements at two spin-locking frequencies and each with different spin-locking times for calibration. Second, injection of ¹⁷O₂-enriched artificial blood can alter blood oxygenation level, resulting in confounding BOLD signal contributions, significantly increasing the uncertainty in determining true MVO₂ values. Third, the T_1ρ measurement is highly sensitive to the static magnetic field (B₀) and shimming, cardiac pulsation, and motion; therefore, controlling the electrocardiogram trigger during imaging acquisitions is critical to reduce imaging signal fluctuations caused by cardiac motion. Additionally, it requires intravenous injection of large amounts of ¹⁷O₂-enriched artificial blood, which is not ideal for potential clinical applications. In contrast, the ¹⁷O-MRS/MRSI technique presented herein overcomes the technical challenges faced by the indirect T_1ρ MRI approach and provides a simple and robust myocardial imaging method that is not only capable of quantitatively imaging MVO₂, but also MBF and OEF from a single imaging measurement.

5 |. CONCLUSIONS

This study demonstrates the feasibility and promise of a noninvasive 3D ¹⁷O-MRSI method for quantitatively imaging myocardial metabolic rate of oxygen consumption, blood flow, and oxygen extraction fraction with a brief inhalation of ¹⁷O₂ and a single imaging measurement. This novel cardiac imaging method is simple, robust, and sensitive to changes in cardiac physiology. The results showed a strong correlation between MVO₂ and MBF under a variety of different physiological conditions, from basal to high workload, suggesting an efficient regulation, controlled by changes in blood perfusion or myocardial oxygen supply, and altered oxygen metabolism that produces adequate ATP energy to meet high cardiac workload demands. With further methodology improvements, the ¹⁷O-MRSI or ¹⁷O-MRI technique promises to provide comprehensive and quantitative measurements of myocardial oxygen metabolism and blood flow, thereby providing oxygen extraction fraction in normal and diseased states, with the potential to translate to human cardiac imaging applications.