In vivo oxygen-17 NMR for imaging brain oxygen metabolism at high field

1. Introduction

The oxygen element is one of the most important components for life on earth because various oxygen containing molecules are present in all levels of biological systems, and oxygen accounts for two thirds of the total human body mass and 90% of the mass of water. However, in vivo oxygen-17 (¹⁷O) NMR has received very little attention compared to other in vivo NMR methodologies, such as ¹H, ¹³C and ³¹P NMR; even though the ¹⁷O NMR signal was first observed in 1951 [1] and utilized since then for many chemical and biochemical applications (see a recent review by Gerothanassis [2, 3] and the cited references therein).

It has been demonstrated that in vivo ¹⁷O NMR can be used to monitor the uptake or washout of an ¹⁷O-labeled exogenous agent (e.g. ¹⁷O-labeled water) for studying tissue perfusion [4–7] or for detecting oxygen-containing metabolites in living species [8–10]. Nevertheless, the most valuable and unique capability of in vivo ¹⁷O NMR is to non-invasively determine the metabolic rate of oxygen in live animals or humans (see [11–13] and references cited therein).

In this review article, we attempt to provide an overview of the methodology background and the present status of in vivo ¹⁷O MR spectroscopy (MRS)/imaging (MRI) approach for imaging the cerebral metabolic rate of oxygen (CMRO₂) and studying the central roles of cerebral oxygen metabolism in brain function. The challenges and potentials of this ¹⁷O-MR based CMRO₂ imaging method will also be discussed.

2. Background

2.1 Importance of oxygen metabolism in brain function and dysfunction

The brain is a highly aerobic organ; it consumes oxygen and glucose extensively in order to generate chemical energy in the form of the adenosine triphosphate (ATP) molecule. A majority of brain energy is used to support the unceasing electrophysiological activities of neurons responsible for inter-neuron transmission and communication throughout the central nervous system. A coupling between neuronal activity and brain energy exists for a wide range of physiological conditions characterized by the highly dynamic change of neuronal activity. This requires a timely efficient balance between ATP demand and supply, which is regulated by a number of crucial biochemical reactions associated with brain metabolisms and neuroenergetics.

Fig. 1 illustrates the key metabolic processes occurring in various sub-cellular compartments and the associated vascular/hemodynamic events of the brain. Oxygen and glucose as the major fuels for brain metabolism are continuously supplied by the circulating blood flow through the capillary bed. Glucose is transported into brain cells and converted to two pyruvate molecules in the cytosol via glycolysis, which are then converted to acetyl-Co A in the mitochondrion and oxidized via the tricarboxylic acid cycle to generate reducing equivalents of NADH and FADH₂. These high energy electron carriers enter the electron transport chain and generate an electrochemical potential gradient across the mitochondrial inner membrane to drive the conversion of adenosine diphosphate (ADP) and inorganic phosphate (Pi) into ATP via oxidative phosphorylation where the electrons are finally transferred to exogenous oxygen and, with the addition of two protons, the final product of water is formed. Under normal physiological conditions with adequate cellular oxygen availability, oxidative phosphorylation comprises ~90% of ATP production [14]. Meanwhile, ATP utilization occurs in the cytosol and results in the reversal conversion of ATP to ADP and Pi with released energy for supporting various neuronal activities.

Fig. 1.

Key metabolic processes occur in various sub-cellular compartments including both mitochondria and cytosol spaces and the associated vascular or hemodynamic events of the brain.

In general, the brain glucose and oxygen consumptions and ATP production are closely coupled in a functional brain as illustrated by Fig. 1. This coupled metabolic reaction chain can be divided into two parts. One is an open system involving the cerebral metabolisms of glucose (from the food chain) and oxygen (from air), resulting in the final products of water and CO₂ that are washed out into the blood stream. Another one is a close system involving extremely efficient cycling between the ATP generation in the mitochondria and utilization in the cytosol inside the cells. These two systems are integrated and work together to maintain normal brain function.

In comparison with the non-oxidative glycolysis, oxidative phosphorylation produces at least 15 times more ATP molecules. Therefore, brain function relies on the ATP energy and the ATP generation relies heavily on the oxygen metabolism in mitochondria. Abnormalities in brain metabolisms, in particular, related to oxidative phosphorylation have been linked to numerous brain disorders and neurodegenerative diseases such as: schizophrenia, Alzheimer’s disease, Huntington’s disease, Parkinson’s disease, mitochondrial dysfunction and aging problems (e.g., [15–19]). There are evidences indicating that the activity of cytochrome oxidase, a key mitochondrial enzyme that catalyzes the reduction of oxygen to form water, is significantly impaired in schizophrenic [17] and Alzheimer’s patients [18, 19]. Also the studies of diseases caused by mitochondrial DNA mutations suggests that a variety of neurodegenerative processes may be associated with defects in mitochondrial oxidative phosphorylation [20, 21]. Therefore, the cerebral metabolic rate of oxygen may provide an early biomarker of pathophysiology change in various brain disorders.

2.2 Concept of studying oxygen metabolism using ¹⁷O NMR

The metabolic processes as illustrated in Fig. 1 can be assessed with various in vivo MR spectroscopy (MRS) techniques. For example, the glucose metabolism has been examined by in vivo ¹H and ¹³C MRS [22, 23], and the ATP metabolism can be studied using in vivo ³¹P MRS in combination with the magnetization transfer method [24–26]. In principle, in vivo ¹⁷O MRS should be able to directly study the oxygen metabolism based on the following simple chemical reaction and the use of ¹⁷O-isotope labeled oxygen gas (¹⁷O₂):

$$4{\text{H}}^{+}+4{\text{e}}^{-}+{{}^{17}\text{O}}_2\to 2{\text{H}}_2{}^{17}\text{O}.$$ [1]

Similar to what has been done with ¹⁵O Positron Emission Tomography (PET) [27, 28], when oxygen gas enriched with the MR detectable ¹⁷O isotope is introduced into the animal or human body, it binds to the hemoglobin in the blood through lung exchange, and then enters the brain via the arteries and blood circulation. The ¹⁷O-labeled oxygen molecules will be metabolized in the brain mitochondrion to produce ¹⁷O-labeled water (H₂¹⁷O) molecules. The production rate of the metabolized H₂¹⁷O water reflects the rate of oxygen metabolism in the brain tissue. Thus, using the ¹⁷O MRS or imaging technique to monitor the dynamic change of the H₂¹⁷O water content in the brain will provide important information regarding the oxygen metabolism of the brain tissue.

2.3 Early attempts of in vivo ¹⁷O NMR studies

The simple concept of utilizing ¹⁷O MR to study oxygen metabolism in living species had been recognized many years ago and attempted by several research labs around the world. In the late 1980s and early 1990s, ¹⁷O-enriched H₂¹⁷O water was used as a potential T₂ (transverse relaxation time) contrast agent for the proton MRI by Hopkins’ group and others for studying tissue perfusion in various animal models [29–33]. Meanwhile, Mateescu et al. demonstrated the in vivo ¹⁷O MRS detection of nascent mitochondrial water in larva and mouse breathing air with ¹⁷O-enriched oxygen gas [34, 35]; Aria et al. explored the feasibility of in vivo ¹⁷O NMR for estimation of cerebral blood flow (CBF) and oxygen consumption in animal models of rat, rabbit and dog [6, 7, 36]; Pekar et al. reported that coarse CMRO₂ images (0.8cc nominal resolution) can be obtained in the cat brain using ¹⁷O NMR imaging and ¹⁷O-enriched oxygen gas [5, 37]; and Fiat et al. examined possible methods for determination of CMRO₂ and CBF in animal brain (and potentially in human brain) using in vivo ¹⁷O MRS/MRI [38–40]. On the other hand, in the middle and late 1990s, Ronen et al. and Reddy et al. suggested that the ¹⁷O-labeled H₂¹⁷O water could be detected by spin-echo proton imaging with ¹⁷O decoupling [41–45] or proton T_1ρ dispersion imaging [46, 47], respectively, and these indirect ¹H-(¹⁷O) methods were used to image the H₂¹⁷O distribution in phantoms and animals.

Although these early efforts had demonstrated the important concept and feasibility of ¹⁷O NMR for assessing brain oxygen metabolism, the technology advancement had been hampered substantially by the limitations in the sensitivity for detecting the metabolically generated H₂¹⁷O water, as well as the practical applicability for general in vivo studies of the oxygen metabolism in animal or human brain.

3. ¹⁷O NMR for studying brain oxygen metabolism

In order to advance the ¹⁷O NMR technology for studying brain oxygen metabolism, it is necessary to thoroughly understand the basic aspects of the ¹⁷O NMR; for instance, the relaxivity and sensitivity of the ¹⁷O signal as well as the factors that influence these properties. In addition, it is important to understand the pros and cons of the different approaches for detecting the H₂¹⁷O signal and find an optimal method for imaging the dynamic changes of the H₂¹⁷O water in vivo. Furthermore, it is essential to develop a reliable method for quantifying the cerebral metabolic rate of oxygen based on ¹⁷O MR measurements of the metabolized ¹⁷O water signal. Finally, it is crucial to establish a simple and completely non-invasive CMRO₂ imaging approach for animal and human applications.

3.1. ¹⁷O NMR properties in a biological system

¹⁷O is a stable isotope of oxygen existing in nature. Unlike the common, abundant form of oxygen (¹⁶O), ¹⁷O is the only oxygen nuclei with a magnetic moment that can be detected by NMR. Different from the nuclei of ¹H, ³¹P and ¹³C used for most in vivo MR applications, ¹⁷O has a spin quantum number of greater than ½ (I = 5/2) and possesses an electric quadrupolar moment. The ¹⁷O nucleus in water can interact with the two protons, resulting in an ¹⁷O signal being a 1:2:1 triplet at extremely low water concentration [48]. However, the ¹⁷O triplets collapse to one single and well-defined H₂¹⁷O resonance peak in a biological sample due to rapid proton exchange and quadrupolar relaxation leading to a large effect of line broadening. This feature simplifies the ¹⁷O MRS pattern for detecting the ¹⁷O water signal in biological systems.

The ¹⁷O natural abundance is only 0.037%, which is almost 30 times lower than that of ¹³C and 2700 times lower than that of ¹H and ³¹P. Moreover, the magnetogyric ratio (γ) of the ¹⁷O, which is proportional to the Larmor frequency, is 7.4 times lower than that of ¹H. These facts result in the lowest NMR receptivity for ¹⁷O spin compared to other spin nuclei commonly used for biomedical research and clinical studies. This low inherent NMR sensitivity has hindered the progress of ¹⁷O NMR for in vivo MR studies especially at relatively low magnetic fields, despite its great potential for providing unique and vital biological information.

3.1.1. Relaxation properties of ¹⁷O in tissue water

One critical aspect of ¹⁷O NMR for in vivo applications is the ¹⁷O relaxivity of water in biological samples. Relaxivity can affect the NMR detection sensitivity and ultimately determines the usefulness and applicability of in vivo ¹⁷O NMR. The ¹⁷O quadrupolar moment can interact with local electric field gradients and the temporal fluctuation in this interaction induced by molecular motion can dominate the ¹⁷O relaxation processes and determine both the longitudinal relaxation time (T₁) and the transverse relaxation time (T₂) [49]. In the case of the water molecule, in which the extreme narrowing limit (i.e., τ_cω ≪ 1, where τ_c is the rotational correlation time and ω is the ¹⁷O Larmor frequency in radian units) is approximately applicable (except for the very small fraction of bound water), the values of ¹⁷O T₁ and T₂ can be approximated by:

$$\frac{1}{T_2}\cong \frac{1}{T_1}=\frac{3{\pi }^2}{10}\left(\frac{2I+3}{I^2(2I-1)}\right)\left(1+\frac{{\eta }^2}{3}\right){\left(\frac{e^{2}Qq}{h}\right)}^2{\tau }_c$$ [2]

where the term (e²Qq/h) is the ¹⁷O quadrupolar coupling constant for bulk water, and η is an asymmetry parameter (0≤ η≤1) [49]. Thus, the ¹⁷O T₁ and T₂ values can be estimated by:

$$\frac{1}{T_2}\cong \frac{1}{T_1}=0.948\left(1+\frac{{\eta }^2}{3}\right){\left(\frac{e^{2}Qq}{h}\right)}^2{\tau }_c.$$ [3]

Because the variables in Eq. [3] are independent upon the magnetic field strength (B₀), ¹⁷O T₁, T₂ as well as apparent T₂ (T₂*) are expected to be insensitive to B₀. Taken the literature values of η=0.7, e²Qq/h = −8.1 MHz and τ_c = 2.7×10⁻¹² s for bulk water at 25°C [50], the estimated ¹⁷O T₂ and T₁ according to Eq. [3] should be approximately 5.1 ms. This estimation indicates that the ¹⁷O relaxation times of water are extremely short, in the range of few milliseconds.

The prediction of equal ¹⁷O T₁ and T₂ values for the water molecules using Eq. [3] relies on the approximation of neglecting the effect of the ¹⁷O-¹H scalar coupling. The actual ¹⁷O water T₂ (or T₂*) value in a biological sample or water solution should be smaller than the ¹⁷O T₁ value because of the combined effects of the ¹⁷O-¹H scalar coupling and the proton exchange on the ¹⁷O transverse relaxation process [51]. The proton exchange rate between H₂¹⁷O and H₂¹⁶O is sensitive to pH. At near-neutral pH, the scalar coupling has a maximal effect for enhancing the apparent ¹⁷O transverse relaxation rate.

The ¹⁷O longitudinal and transverse relaxation times of the natural abundance water in the rat brain have been explicitly measured and compared at field strengths of 4.7T versus 9.4T [52]. The relaxation times were found to be field independent (T₂=3.0 ms, T₂*=1.8 ms and T₁=4.5 ms at 4.7T versus T₂=3.0 ms, T₂*=1.8 ms and T₁=4.8 ms at 9.4T) [52]. These experimentally measured ¹⁷O relaxation times are in line with the predicted values according to Eq. [3] and the reported values in the literature [38, 53]. Recent experimental evidence has indicated that the field-independence of the ¹⁷O relaxivity can hold at much higher magnetic fields [54].

3.1.2. Pros and cons of extremely short ¹⁷O-water relaxation times

The very short ¹⁷O T₁ values of the brain H₂¹⁷O (several milliseconds) [38, 52] allow rapid NMR signal acquisitions at a given sampling time, thus, gaining signal-to-noise ratio (SNR) per unit time. The repetition time for acquiring the ¹⁷O signal can be pushed to as short as tens of milliseconds. The major constraint is the potential concern of the specific absorption rate (SAR: a measure of the rate of absorption of RF energy in the body) allowed by FDA regulation.

The extremely short ¹⁷O T₂* of the brain H₂¹⁷O (~2 ms) results in a broadening of the water resonance peak to a line width of 100–200 Hz in vivo and a substantial SNR reduction. It is crucial to minimize the delay (or echo time) between the ¹⁷O spin excitation pulse and NMR signal sampling in order to avoid a substantial loss of the ¹⁷O signal due to the rapid T₂* decay. The linewidth of the ¹⁷O resonance peak of H₂O is relatively insensitive to the B₀ inhomogeneity (hence shimming quality) because of the intrinsically broad linewidth and the much lower ¹⁷O magnetogyric ratio (7.4 times lower than that of ¹H). This fact implies that the requirement for B₀ homogeneity either in the bare magnet or together with room temperature shim compensation is considerably less stringent for in vivo ¹⁷O NMR compared to in vivo ¹H, ³¹P or ¹³C NMR.

3.1.3. Advantage of ¹⁷O NMR at high/ultrahigh magnetic field

One of the most important advantages provided by high/ultrahigh magnetic fields is the potential gain in NMR sensitivity. This is particularly crucial for in vivo ¹⁷O NMR, where the inherent NMR sensitivity is extremely low. For a magnetic nucleus, the optimal SNR of the NMR signal acquired within a unit time at a given field strength depends on T₁, T₂*, B₀ and the RF coil quality factor (Q) according to [55–58]:

$$\text{SNR}(\text{per}\text{unit}\text{acqusition}\text{time})\propto {\text{B}}_0^{\beta }\sqrt{\frac{{QT}_2^{\ast }}{T_1}}.$$ [4]

The parameter β was suggested to be approximately 7/4 based on theoretical predictions [56, 58]. Unlike the water proton spins in biological tissues, which are characterized by longer T₁ and shorter T₂ (or T₂*) with increased field strength, the field independence of ¹⁷O relaxivity implies that the ¹⁷O-water sensitivity gain at higher fields is not compromised by the relaxation times. Although the short T₂* (or broad linewidth) of the ¹⁷O resonance peak in H₂¹⁷O leads to an effective reduction in the ¹⁷O NMR sensitivity, this reduction can be partially compensated by the extremely short ¹⁷O T₁ (< 5 ms in the brain), allowing rapid signal averaging [38, 52]. Therefore, it is possible to achieve a large sensitivity gain for in vivo ¹⁷O NMR at high/ultrahigh fields.

The in vivo ¹⁷O NMR sensitivity has been quantitatively compared at field strengths of 4.7T versus 9.4T [52]. The striking finding from this study, as shown in Fig. 2, is the consistent observation of approximately four-fold SNR gain at 9.4T compared to 4.7T indicating an approximated 7/4^th power dependence of ¹⁷O SNR on B₀ as predicted by the NMR theory [52, 56, 58]. These results demonstrate the significant advantage provided by high field strength for the direct detection of ¹⁷O NMR signal. The trend for increasing ¹⁷O NMR sensitivity is likely to hold much beyond the field strength of 9.4T [54]. Compare to in vivo ¹H, ³¹P or ¹³C NMR, in vivo ¹⁷O NMR is likely to benefit the most from the increasing field strength in terms of the NMR sensitivity gain.

Fig. 2.

(A) One-dimensional SNR profiles of ¹⁷O-water signal in the rat brain at 4.7T and 9.4T, and the SNR ratio between 9.4T and 4.7T; (B) Single voxel ¹⁷O-MR spectrum of H₂¹⁷O signal obtained from rat brain at 4.7T and 9.4T, respectively (total acquisition time of 15 s, and nominal voxel size of 16 μl. Adapted from Zhu et al. MRM 2001; 45: 543–549.

3.1.4. In vivo NMR invisibility of ¹⁷O₂

In contrast to the ¹⁵O-PET approach which is unable to distinguish the signals emitted by the ¹⁵O atoms in the ¹⁵O₂ molecules from those in the metabolically generated H₂15O molecules; the ¹⁷O resonance peak of the ¹⁷O₂ molecules, when bound to hemoglobin in the blood, is extremely broad because of very slow rotational motion of the large HbO₂ complex, and very difficult to detect with conventional in vivo ¹⁷O NMR approaches. Saturation transfer electron paramagnetic resonance studies have shown that the τ_c value for the rotational motion of the hemoglobin molecule is 2·10⁻⁸ s in solution and increases to 8·10⁻⁶ s when the hemoglobin molecule is encapsulated within the erythrocyte [59]. This τ_c value is approximately 10⁶ times slower than that of the free water. Such slow rotational motion leads to extremely fast ¹⁷O T₂ relaxation according to Eq. [2] and renders the ¹⁷O₂ molecule bound to hemoglobin invisible for in vivo ¹⁷O NMR detection.

The ¹⁷O₂ molecule while it is in the gas phase or dissolved in water is strongly paramagnetic due to its two unpaired electrons, and hence is undetectable in conventional NMR measurement because of the strong dipolar coupling between the electrons and the ¹⁷O nucleus. Thus, the direct in vivo ¹⁷O NMR approach will detect only H₂¹⁷O but not ¹⁷O₂ in the biological sample.

This unique MR specificity for detecting ¹⁷O-labeled water significantly simplifies the methodology for measuring and quantifying cerebral metabolic rate of oxygen (CMRO₂) because all ¹⁷O-labeled components other than the metabolically generated H₂¹⁷O can be ignored, leading to a simple quantification scheme as illustrated in Fig. 3 [60].

Fig. 3.

Schematic illustration of a “complete model” describing three parallel processes of the ¹⁷O-labeled metabolic water (H₂¹⁷O) occurring in the brain when the ¹⁷O-labeled oxygen gas molecules are introduced via an inhalation. In this model, only the metabolic H₂¹⁷O is considered because the ¹⁷O-labeled O₂ is invisible by in vivo ¹⁷O NMR. C_a(t), C_b(t) and C_v(t) stand for the H₂¹⁷O concentration in arteriole, brain tissue and venule, respectively, as a function of the ¹⁷O₂ inhalation time.

3.2. Influence of ¹⁷O-¹H scalar coupling on ¹H relaxation of water

The existence of an ¹⁷O spin in the tissue water, either in natural abundance or at elevated enrichment levels, will affect the ¹H relaxation properties of the tissue. Half a century ago, Meiboom studied the proton chemical exchange in ¹⁷O-enriched water solution at pH of neutral range [51]. It was found that the spin-spin coupling (i.e. scalar coupling) between the ¹H and the ¹⁷O spins could shorten the proton transverse relaxation time or the rotating frame spin-lattice relaxation time (T_1ρ) but not the longitudinal relaxation time (T₁), and such scalar interaction of ¹⁷O-¹H was modulated by the fast proton chemical exchange between the H₂¹⁷O and H₂¹⁶O molecules. Meiboom’s finding provided a mechanism for indirect detection of the H₂¹⁷O signal using ¹H MR imaging. This indirect ¹H-(¹⁷O) MR approach was anticipated to better inherent NMR sensitivity than that of the direct ¹⁷O MR approach [41].

3.2.1. Effect of ¹⁷O-¹H scalar coupling on proton T₂ relaxation of tissue water

The influence of the ¹⁷O-¹H scalar coupling on the ¹H T₂ value (T_2,H) of the tissue water can be quantitatively described by [29, 41, 42, 51, 61]:

$$\frac{1}{T_{2,H}}=\frac{1}{T_{2,H}^{(16)}}+P(\frac{1}{T_{2,H}^{(17)}}-\frac{1}{T_{2,H}^{(16)}})\approx \frac{1}{T_{2,H}^{(16)}}+\frac{35}{12}P\tau J^2,$$ [5]

where a ¹H chemical shift difference between H₂¹⁷O and H₂¹⁶O is neglected; $T_{2,H}^{(16)}$ and $T_{2,H}^{(17)}$ are the proton transverse relaxation times of H₂¹⁷O and H₂¹⁶O, respectively; P is the molar fraction of H₂¹⁷O and is equivalent to the ¹⁷O enrichment fraction; τ is the characteristic proton exchange lifetime in H₂¹⁷O and J is the ¹⁷O-¹H scalar coupling constant. The relation described by Eq. [5] provides the basis for assessing the fractional content of H₂¹⁷O water (i.e., P in Eq. [5]) through the change in proton T₂.

It was experimentally verified that the proton transverse relaxation rate indeed linearly correlates with the H₂¹⁷O concentration in biological solutions up to 5% enrichment, whereas the proton longitudinal relaxation time was not significantly affected by the ¹⁷O enrichment [29]. Therefore, ¹⁷O-labeled H₂¹⁷O can be used as an exogenous and/or endogenous contrast agent, respectively, for cerebral blood flow (CBF) and CMRO₂ measurements through measuring the change in the T₂-weighted ¹H MRI signal using a spin-echo sequence, for instance.

Several early studies observed proton signal reduction in the brain after the introduction of ¹⁷O-labeled water [7, 29–32, 62], and its change can be linked to cerebral blood perfusion at different physiological conditions in normal and ischemic brain (e.g., [31]). The time courses of the signal changes were successfully applied to quantify and image CBF for a wide range of CBF values induced by hypercapnia and hypocapnia in animals [7, 32]. Subsequently, this indirect ¹⁷O detection approach was applied for imaging the signal change of the H₂¹⁷O metabolized from the ¹⁷O-labeled oxygen gas [36, 63]. However, all these early studies were conducted in a qualitative manner and it was difficult to provide an absolute concentration of the ¹⁷O-labeled water in the brain. Though this limitation should not be a problem for quantifying CBF in which only the relative concentration of H₂¹⁷O water in the brain is required, it does pose a hurdle for quantifying CMRO₂.

One improved method for quantifying the H₂¹⁷O tracer concentration is to introduce another proton spin-echo MRI in the presence of ¹⁷O decoupling during the echo time (TE) and/or acquisition time [41–45]. The ¹⁷O decoupling can abolish the interaction of the ¹H-¹⁷O scalar coupling, and ultimately, suppress its effect on the water proton T₂ relaxation and increase the T₂ value. The difference of the proton spin-echo signals of H₂¹⁷O water in the presence and absence of ¹⁷O decoupling provides a simple connection with the H₂¹⁷O content. The absolute H₂¹⁷O concentration can be calibrated by additional paired measurements of natural abundance H₂¹⁷O with a known concentration (i.e., P = 0.037%) before introducing the H₂¹⁷O tracer. This indirect ¹⁷O approach with ¹⁷O decoupling was tested with phantom solutions and tissue models [41, 43–45]; it was also examined in the rat brain under normal physiological condition [42] and during focal cerebral ischemia [64] by injection of ¹⁷O-labeled water tracer into the rat body. The results indicate that when an adequate amount of metabolic H₂¹⁷O is present in the brain, this indirect in vivo ¹⁷O approach could potentially be used for quantifying and imaging CMRO₂.

3.2.2. Effect of ¹⁷O-¹H scalar coupling on proton T_1ρ relaxation of tissue water

Another alternative indirect ¹⁷O approach is to detect H₂¹⁷O by using the proton T_1ρ dispersion imaging method [46, 47, 65]. This method applies single proton radiofrequency (RF) channel for acquiring T_1ρ-weighted proton images with two different spin locking RF powers. Although the mechanism underlying the T_1ρ contrast in biological samples is not fully understood, the major contribution is likely to be from proton exchange [66], which can be potentially linked to the H₂¹⁷O concentration through ¹⁷O-¹H scalar coupling and the chemical exchange between H₂¹⁷O and H₂¹⁶O. Such an exchange, however, may not be the only process occurring in biological samples, and this complicates the T_1ρ method and quantification for determining the H₂¹⁷O concentration. An example of this complication is the observation that T_1ρ contrast changes in the animal brain during ischemia even without ¹⁷O-labeled water tracer [67, 68]. Another complication is the difficulty of obtaining prior information on the intrinsic T_1ρ dispersion of tissue, which is needed for absolute quantification. Nevertheless, this method has been successfully applied to determining CBF where the absolute H₂¹⁷O concentration is not required [69], and to imaging the CBF changes in tumors [70]. Recently, it has also been attempted for estimating the CMRO₂ in the animal brain [71–73].

3.2.3. Pros and cons of proton MRI for indirectly detecting ¹⁷O-labeled metabolic water

The major advantages of indirect ¹H-(¹⁷O) approaches for detecting ¹⁷O-labeled water in brain tissue is the high sensitivity of the proton signal and the ability to apply conventional MRI acquisition and data processing methods. In addition, the proton T_1ρ dispersion imaging approach requires only a single proton RF channel and can be implemented on a clinical MRI scanner.

However, caution should be exercised when absolute quantification of the H₂¹⁷O concentration in a biological sample is required. Both ¹⁷O-¹H scalar coupling and the chemical exchange between H₂¹⁷O and H₂¹⁶O, the mechanisms responsible for the indirect ¹⁷O detection, are all sensitive to many physiological parameters such as pH and temperature. This difficulty is evident from the experimental observations that equal concentrations of H₂¹⁷O tracer do not produce the same magnitude of T₂ change (or T₂-weigthed proton signal) in different physiological environments [29], and the detected change can completely disappear when pH is shifted away from neutral [64].

Furthermore, it should also be noted that although the intrinsic proton signal of the tissue water offered by indirect ¹H-(¹⁷O) approach is much higher than the ¹⁷O signal measured by the direct ¹⁷O approach, the reproducibility or reliability of the ¹⁷O-water signal detection in consecutively acquired datasets is far more crucial than the absolute signal (or SNR) acquired in a single dataset. This is because the ¹⁷O-MR based CMRO₂ imaging approach relies on measuring the small dynamic changes of the metabolically generated H₂¹⁷O and their spatial distribution. In addition, the actual ¹H signal that is relevant to the tissue H₂¹⁷O content could be much smaller than the total available ¹H signal. For example, it has been shown that a 10% signal increase in the T₂-weighted ¹H MRI corresponded to about 0.45% of H₂¹⁷O content in the rat brain, which is over twelve times that of the natural abundance H₂¹⁷O level [42]. Therefore, the proton signal changes due to the metabolically generated H₂¹⁷O water in the T₂- or T_1ρ-weighted MR imaging, which is expected to be in the range of 1%, will likely be compromised by the signal fluctuation caused by the physiological noise (e.g., respiration or pulsation) or scanner instability [13].

Another potential technical limitation posed by both ¹⁷O-decoupled and T_1ρ-based ¹H MRI approaches is the requirement of relatively large RF power either for ¹⁷O decoupling or for the proton spin locking, in particular, for human applications at high magnetic fields.

3.3. Theory and quantification of CMRO₂ based on in vivo ¹⁷O MRS/MRI approach

3.3.1. Theory and quantification model

As illustrated in Fig. 3, the dynamic change of the metabolically generated H₂¹⁷O concentration in the brain during an ¹⁷O₂ inhalation is affected by three parallel processes: (i) cerebral oxygen utilization for generating the metabolic H₂¹⁷O in the brain tissue; (ii) cerebral blood perfusion resulting in H₂¹⁷O washout from the brain, and (iii) blood recirculation bringing the metabolically generated H₂¹⁷O in the entire body back to the brain. All contributions from these three processes have to be considered for quantifying CMRO₂. Based on the Kety - Schmidt theory [74], the mass balance of the ¹⁷O-isotope labeled H₂¹⁷O in the brain tissue during an ¹⁷O₂ gas inhalation can be derived as [5, 39, 60, 75]:

$$\frac{{dC}_b(t)}{dt}=2\alpha (t)f_1{\mathit{CMRO}}_2+f_2\mathit{CBF}(C_a(t)-C_v(t)),$$ [6]

where C_a(t), C_b(t) and C_v(t) are the metabolic H₂¹⁷O concentrations in excess of the natural abundance H₂¹⁷O concentration in the arterial blood, brain tissue and venous blood; respectively, as a function of ¹⁷O₂ inhalation time (t, unit = minute); α(t) is the ¹⁷O enrichment fraction of the oxygen atoms in the inhaled ¹⁷O₂ gas which could vary with inhalation time; and the factor of 2 accounts for the fact that two H₂O molecules are formed from one O₂ molecule through the oxidative metabolism according to Eq. [1]. Two unit conversion factors, f₁=1.27 and f₂=1.05, are used to achieve consistency of units among all parameters used in Eq. [6] [37, 60, 75].

The natural abundance H₂¹⁷O concentration can be used as an internal reference to calibrate the absolute values of C_b(t), C_a(t) and C_v(t) with preferred units of μmol/(g brain water) for C_b(t), μmol/(g blood water) for C_a(t) and C_v(t); and leads to the CMRO₂ unit of μmole/min/(g brain tissue).

Therefore, the CMRO₂ values can be precisely calculated by solving the linear differential Eq. [6] if the parameters of C_b(t), α(t), CBF, C_a(t) and C_v(t) are known or can be measured.

3.3.2. CMRO₂ quantification for small animal models

The methods for quantifying CMRO₂ in the small animal case were first examined by Pekar and Fiat et al. [5, 37, 39, 40, 76]. The complete model of CMRO₂ quantification in the small animal was clearly demonstrated and established by Zhu et al. for fast imaging of CMRO₂ in rat brain within a few minutes of ¹⁷O₂ inhalation at high field [75]. In this comprehensive study, all parameters involving the oxygen metabolism and perfusion of the brain tissue as shown in Eq. [6] were experimentally determined via independent and concurrent ¹⁷O MR measurements in the rat brain at 9.4T [75].

For a small animal such as a rat, due to its fast respiration and high heart rate, the labeled ¹⁷O₂ gas once introducing into the body, will quickly replace the regular ¹⁶O₂ gas to produce H₂¹⁷O water. Thus, the ¹⁷O enrichment fraction can be approximated as a time independent constant, i.e. α(t) ≈α, and the transition time for ¹⁷O₂ to replace the ¹⁶O₂ gas is negligible compare to the total ¹⁷O₂ inhalation period (≥ few minutes). If one assumes that the water in the brain tissue is in equilibrium with water in the venous blood, then f₂C_v(t) = C_b(t)/λ where λ is the brain/blood partition coefficient (≈ 0.90) with the unit of (ml blood)/(g brain tissue) [77]. Substituting this relation and introducing two new correction factors (n and m) into Eq. [6] leads to

$$\frac{{dC}_b(t)}{dt}=2\alpha f_1{\mathit{CMRO}}_2+\mathit{mCBF}(f_{2}Ca(t)-\frac{{nC}_b(t)}{\lambda }).$$ [7]

The correction factor m accounts for the water permeability restriction across the blood-brain burrier (BBB) [78]; and n accounts for the permeability restriction when the metabolically generated H₂¹⁷O molecules inside the mitochondria crosses the mitochondrial membranes [60, 75]. Both m and n depend on the CBF [60, 75]. The function of C_a(t) (or artery input function) is determined by the total metabolic H₂¹⁷O generated in all aerobic organs of living body. It approximates as a linear function of ¹⁷O₂ inhalation time (i.e., C_a(t)≈At, where A is a constant) [5, 6, 60, 75]. Thus, the solution for Eq. [7] becomes:

$${\mathit{CMRO}}_2(t)=\frac{\left\{\left[\frac{C_b(t)-\frac{{Af}_2{\lambda }^2}{{mn}^2\mathit{CBF}}(\frac{\mathit{mnCBF}}{\lambda }{te}^{-\frac{\mathit{mnCBF}}{\lambda }t}+e^{-\frac{\mathit{mnCBF}}{\lambda }t}-1)}{1-e^{-\frac{\mathit{mnCBF}}{\lambda }t}}\right]-\frac{{Af}_2\lambda t}{n}\right\}}{\frac{2\alpha \lambda f_1}{\mathit{mnCBF}}}.$$ [8]

According to this equation, the CMRO₂ value at each data point measured at different inhalation time (t) can be precisely calculated using the experimentally measured CBF, A and n values, C_b(t) time courses and other known constants (f₁, f₂, m, α and λ) [13, 60, 75].

Fig. 4 summaries the multiple in vivo ¹⁷O MR measurements performed on anesthetized rat brains for imaging and quantifying CMRO₂ [13, 75]. The CBF measurement was performed via bolus injection of a small amount of ¹⁷O-enriched H₂¹⁷O into one internal carotid artery and monitoring the washout process of the H₂¹⁷O tracer in the animal brain using 3D ¹⁷O chemical shift imaging (CSI) [52]. Fig. 4A demonstrates the stacked plots of H₂¹⁷O spectra acquired from a single voxel of a 3D ¹⁷O CSI data set in a representative rat before and after the H₂¹⁷O bolus injection. The peak height of the H₂¹⁷O spectra shows an exponential decay and its decay rate determines the CBF value in the CSI voxel [13, 52, 75]. The crucial step for CMRO₂ measurements is to monitor and image the dynamic changes of the metabolic H₂¹⁷O content in the brain (i.e., C_b(t)) during an inhalation of ¹⁷O₂ gas. Fig. 4B illustrates the stacked plots of ¹⁷O spectra of cerebral H₂¹⁷O from one representative CSI voxel acquired before, during and after a 2-minute inhalation of ¹⁷O₂ [75]. It indicates excellent ¹⁷O NMR sensitivity for detecting the cerebral H₂¹⁷O signal and its change during the inhalation; and the approximately linear increase of brain H₂¹⁷O during a short ¹⁷O₂ inhalation is evident, and the slope is tightly coupled to CMRO₂. The arterial input function C_a(t) was measured in vivo by an implanted ¹⁷O RF coil [79] wrapped around a carotid artery. Fig. 4C illustrates the implanted ¹⁷O RF coil, the natural abundance H₂¹⁷O signal detected only from the rat carotid blood and the C_a(t) time course measured during a two-minute inhalation of ¹⁷O₂ [75]. The experimental results show an approximately linear relation between the arterial H₂¹⁷O concentration and the ¹⁷O₂ inhalation time, and the linear regression of C_a(t) gave the value of the constant A required by Eq. [7] and Eq. [8]. Finally, the ratio between the decay rates of H₂¹⁷O signal measured after the cessation of ¹⁷O₂ inhalation (see Fig. 4B) versus that after a H₂¹⁷O bolus injection (see Fig. 4A) gave the value of the constant n reflecting the H₂¹⁷O permeability restriction across the mitochondrial membranes [60, 75].

Fig. 4.

Schematic diagram showing the multiple in vivo ¹⁷O measurements at 9.4T for determining CMRO₂ using the complete model according to the mass balance equation of Eq. [7] which links C_b(t), C_a(t), CBF and n with CMRO₂. To simplify the equation, three known constants of 2αf₁, mf₂ and m/λ used in Eq. [7] are replaced by k₁, k₂ and k₃, respectively. (A) Stacked plot of the ¹⁷O spectra of cerebral H₂¹⁷O tracer from one representative voxel as indicated by the circle in the anatomical brain image (low center insert). The spectra were acquired before and after a bolus injection of H₂¹⁷O for CBF measurements. (B) Stacked plot of the ¹⁷O spectra of the metabolic H₂¹⁷O from the same voxel acquired before (natural abundance), during (as indicated by the gray bar under the stacked plot) and after a 2-minute ¹⁷O₂ inhalation. (C) Measurement of C_a(t) by using an implanted ¹⁷O RF coil (the left insert). The middle insert illustrates an ¹⁷O spectrum of natural abundance H₂¹⁷O obtained from the rat carotid artery blood with the implanted coil before ¹⁷O₂ inhalation. The right insert shows the time course of C_a(t) (circle symbol) and C_b(t) from a representative 3D ¹⁷O CSI voxel (square symbol) in the same rat during the ¹⁷O₂ inhalation. Finally, the ratio between the ¹⁷O signal decay detected after a bolus injection of H₂¹⁷O (see Fig. 4A) versus the ¹⁷O signal decay detected after the cessation of ¹⁷O₂ inhalation (see Fig. 4B) gives the constant of n. Adapted from Zhu et. al. PNAS 2002; 99: 13194–13199.

The CBF measurement can be performed concurrently with the CMRO₂ measurement on the same animal while C_a(t) (see Fig. 4C) and C_b(t) (see Fig. 4B) measurements can be conducted simultaneously with the configuration of dual ¹⁷O RF coils and receivers [79]. The values of C_b(t), CBF and n measured from each ¹⁷O MRS imaging (MRSI) voxel and the value of A measured from each ¹⁷O inhalation measurement in the same animal as demonstrated in Fig. 4 can be used to calculate the absolute CMRO₂ value as a function of inhalation time according to Eq. [8]. Fig. 5A shows one example of CMRO₂ time course from a representative ¹⁷O-CSI voxel with a temporal imaging resolution of 11 seconds [75]. It is evident that the CMRO₂ values are independent of the ¹⁷O₂ inhalation time if the first two CMRO₂ values characterized with relatively large fluctuations are excluded. These CMRO₂ values were averaged for improving measurement accuracy. The same procedure and calculation can be applied to all ¹⁷O CSI voxels for generating 3D CMRO₂ images in the rat brain [12, 75]. Fig. 5B demonstrates three adjacent CMRO₂ images in the coronal orientation from a representative rat brain. The averaged CMRO₂ and CBF values in the rat brains anesthetized with α-chloralose were found to be 2.19±0.14 μmol/g/min and 0.53±0.07 ml/g/min (n=7), respectively [75]. These results are consistent with the literature reports using other independent techniques under similar physiological condition [80, 81].

Fig. 5.

(A) Plot of the calculated CMRO₂ values using the complete model as described by Eq. [8] as a function of ¹⁷O₂ inhalation time. (B) Three-dimensional coronal CMRO₂ images of rat brain measured by in vivo ¹⁷O MRS approach during a 2-minute ¹⁷O₂ inhalation at 9.4 Tesla. Adapted from Zhu et. al. PNAS 2002; 99: 13194–13199.

3.3.3. CMRO₂ quantification in humans

Unlike small animals, quantification of CMRO₂ in humans faces serious challenges. The human body size, lung capacity and respiration rate as well as blood circulation speed are drastically different from those in the small animal. It is expected that the exchange process between non-labeled and inhaled ¹⁷O-labeled oxygen gases in a human lung will be much slower compared to small animals such as a rat or mouse. Moreover, a much longer blood circulation time through the human body could further slow down the binding process of inhaled ¹⁷O-labeled oxygen to the hemoglobin in the blood stream. Thus, the ¹⁷O fractional enrichment of the oxygen gas in human artery blood (i.e., the α(t) term in Eq. [6]) will take much longer to reach a steady-state level (i.e., the ¹⁷O enrichment α of the inhaled ¹⁷O₂ gas). For this reason, the CMRO₂ quantification model that worked well for the rat brain [75] failed to provide acceptable CMRO₂ values for a human study in which a short inhalation of ¹⁷O₂ gas (2–3 minutes) was employed [82].

Recently, Atkinson and Thulborn proposed a three-phase model for quantifying CMRO₂ in the human brain based on in vivo ¹⁷O and ²³Na MR imaging data obtained at 9.4T [83]. In this model, the dynamic change of the H₂¹⁷O water in brain tissue was separated into three-phases: i.e., prior, during and after inhalation of ¹⁷O₂ gas; two rate constants K_L and K_G were utilized to represent the loss and gain of the H₂¹⁷O water within the imaging voxel, respectively; and a mass balance equation similar to Eq. [6] was used to describe the amount of the ¹⁷O-labeled water in each voxel at each phase where the brain mass of the voxel was computed from co- registered ²³Na MRI data. The CMRO₂ as well as the rate constants K_L and K_G values for each imaging voxel were determined by performing a least-square fit of the dynamic H₂¹⁷O data for all three-phases. The key component of this model for applying to the CMRO₂ quantification in human is that it considered the transition of the ¹⁷O fractional enrichment of the ¹⁷O₂ gas in arterial blood (i.e. the α(t) term of Eq. [6]). The transitions of α(t) from zero to α during the inhalation phase and from α to zero during washout phase after the inhalation were estimated based on the parameters of the pulmonary arteriovenous difference fraction (F_A-V) and the mean blood circulation time (T_C), in which their values were approximated from the literature reports [83].

Although there are still many uncertainties in the CMRO₂ quantification model for human application, especially for determining and validating the fractional enrichment of the ¹⁷O₂ gas α(t), the work of Atkinson and Thulborn does provide a forward step towards the quantitative study of the oxygen metabolism in human brain using the in vivo ¹⁷O MR imaging approach at high/ultrahigh field.

3.4. Establishing a robust and noninvasive ¹⁷O MR method for Imaging CMRO₂

3.4.1. Methods for imaging dynamic H₂¹⁷O water content of the brain tissue

The most crucial measurement for determining the CMRO₂ value is to monitor the dynamic change of the H₂¹⁷O water content in brain tissue. Therefore, establishing a robust CMRO₂ imaging approach relies on the ability to reliably image the dynamic change of the H₂¹⁷O signals in the brain tissue with reasonable spatial and temporal resolution.

For the indirect ¹H-(¹⁷O) detection approach, conventional T₂-weighted spin-echo or T_1ρ- weighted MR imaging technique can be used to acquire the proton signal change related to the variation in H₂¹⁷O water content (see detailed discussion in Section 3.2.). For direct ¹⁷O-MR detection approach, however, the conventional MR imaging technique is no longer suitable for imaging the H₂¹⁷O signal because the T₂ (or T₂*) relaxation time of the H₂¹⁷O is extremely short, leading to severe signal loss or even disappearance if the echo time of the MR imaging sequence is relatively long compare to the transverse relaxation time of the ¹⁷O-water (in the few millisecond range, see details in Section 3.1.).

So far, two MR imaging approaches with ultra-short echo time capability have been used for directly imaging the ¹⁷O-water signal. One approach applies the flexible twisted projection [84] or a density-adapted 3D radial pulse sequence [85] commonly used for acquiring sodium MRI data; and the other approach images the H₂¹⁷O signals directly using the 3D chemical shift imaging (CSI) technique [86, 87]. Fig. 6 illustrates a typical 3D CSI sequence (Fig. 6A) used for the in vivo ¹⁷O MR measurement and an example of the ¹H anatomic image as well as 3D ¹⁷O-CSI data of the natural abundance H₂¹⁷O signal obtained from a representative rat brain (Fig. 6B). These in vivo ¹⁷O-MR image data were acquired with approximately 0.1 ml of spatial resolution (nominal resolution: ~40μl) and total acquisition time of ~11 seconds at 9.4T. Regardless of which imaging sequence is chosen, the sensitivity and reliability of the ¹⁷O-water signal obtained with the direct ¹⁷O imaging approach depends in large extent on the echo time used for the imaging assuming comparable spatial and temporal resolutions are applied.

Fig. 6.

(A) 3D chemical shift imaging sequence; and (B) 3D ¹⁷O CSI data of natural abundance H₂¹⁷O (top row) and corresponding ¹H anatomical images (bottom row) of rat brain acquired at 9.4T. The RF ¹⁷O surface coil positions and cross sections are indicated in the images.

3.4.2. Feasibility of establishing a noninvasive CMRO₂ imaging approach

As described earlier, the major technical limitation of the complete model for noninvasively determining CMRO₂ is the requirement of invasive measurements (e.g., CBF, C_a(t), n). This will significantly limit the potential of this in vivo ¹⁷O neuroimaging approach for broad biomedical applications, especially in humans. Thus, it is crucial to examine the feasibility of developing a completely noninvasive ¹⁷O approach for imaging CMRO₂. Attempts have been made to simplify the experimental procedures and the models for determining CMRO₂ based on a number of approximations [5, 37, 39, 40, 60, 76].

One of these attempts, in which the invasive measurements could be eliminated completely by using the simplified model based on expanding C_b(t) as a polynomial [60],

$${\text{C}}_{\text{b}}(\text{t})={\text{a}}_1\text{t}+{\text{a}}_2{\text{t}}^2+{\text{a}}_3{\text{t}}^3+\cdots .$$ [9]

In this expansion, the first-order (or linear) coefficient of a₁ is directly proportional to CMRO₂ according to the following equation |[60]

$$\mid {\text{CMRO}}_2=\frac{{\text{a}}_1}{2\alpha {\text{f}}_1},$$ [10]

where α and f₁ are known constants. Thus, using this simplified model, only the time course of C_b(t) measured noninvasively by in vivo ¹⁷O MR imaging is needed and it can be fitted to the polynomial function in Eq. [9] to calculate the linear coefficient of a₁, and ultimately determining CMRO₂ according to Eq. [10]. For practical applications with adequate SNR, a quadratic polynomial function usually provides a good approximation for fitting the time course of C_b(t) with a moderated fitting error |[60]. It has been demonstrated that the CMRO₂ value obtained based on the complete model with invasive procedures has no statistical difference from that based on the simplified model and quadratic function fitting where only a single noninvasive measurement of C_b(t) is required |[60]. Moreover, the results also indicate that the linear fitting of C_b(t) could provide a good approximation for determining CMRO₂ in the rat brain when the ¹⁷O₂ inhalation time is relatively short (e.g., 2 minutes) |[60]. In another study, the simplified model was examined for determining CMRO₂ under varied physiological conditions |[88]; and the CMRO₂ results obtained with the complete model were compared with the simplified model using linear fitting of C_b(t) under normothermia (37°C) and hypothermia (32°C) condition, which is a well known factor leading to significant suppression of both CBF and CMOR₂. Fig. 7 demonstrates an excellent consistency of the CMRO₂ results between the complete and simplified models for either the voxel-based comparison (Fig. 7A) or the averaged CMRO₂ comparison (Fig. 7B) at both brain temperatures |[88]. The comparison results reveal the validity of the simplified ¹⁷O approach for imaging CMRO₂ applicable for small animal work across a wide physiological range

Fig. 7.

(A) Voxel based CMRO₂ calculation and comparison using the completed and simplified models from a representative rat (total voxel number used was 224 for 32°C and 254 for 37°C, voxel size = 75μl ). (B) Averaged CMRO₂ values in the same rat brain at normothermia (37°C) and hypothermia (32°C) condition, calculated with simplified and completed model, respectively. Adapted from Zhu et. al. JCBFM 2007; 27(6): 1225–1234.

The ability to non-invasively image CMRO₂ in human brain is even more crucial. However, the above mentioned simplified method by linear or quadratic fitting of the C_b(t) data is no longer appropriate because of the time dependent ¹⁷O fractional enrichment of the ¹⁷O₂ gas (see details in Section 3..3.3) in humans. The quantification method based on the three-phase model has the potential for non-invasively mapping CMRO₂ in humans after careful validation and improvement |[83]. It is anticipated that additional technical developments which further advance the in vivo ¹⁷O methodology could ultimately provide the simplest and completely non-invasive ¹⁷O neuroimaging approach for imaging CMRO₂ in both animal and human brains.

3.4.3. Reliability and reproducibility of the CMRO₂ measurement

Another merit of the ¹⁷O CMRO₂ imaging approach is its ability for performing repeated CMRO₂ imaging measurements with a short interval between two measurements. This is because the cerebral H₂¹⁷O concentration can quickly reach a new and steady level after the cessation of a brief ¹⁷O₂ inhalation, which allows repeated CMRO₂ measurements in the same subject and experimental session (see Fig. 8A for an example). Fig. 8B shows the excellent reproducibility of repeated CMRO₂ measurements in five rats (1^st measured CMRO₂ = 2.26 ± 0.18; 2^nd measured CMRO₂ = 2.20 ± 0.14 μmol/g/min; CMRO₂ ratio between the 1^st and 2^nd measurements = 1.03 ± 0.05; n=5) where CMRO₂ values were determined solely from the dynamic change of the ¹⁷O-water signals |[88]. The results demonstrate the robustness and reliability of the simplified in vivo ¹⁷O NMR approach for noninvasively and rapidly imaging CMRO₂ repeatedly in the small brain of a rat. This capability is particularly valuable for studies aiming at CMRO₂ changes induced by physiological or pathological perturbations in which multiple measurements are required under different conditions (e.g., control versus stimulation for brain function study). Therefore, it is likely that, at least in small animal brains, the combination of the simplified model and ultrahigh-field in vivo ¹⁷O MRS may provide an alternative neuroimaging modality for studying the central role of oxidative metabolism in brain function and neurological diseases |[12, 13].

Fig. 8.

(A) Stacked plots of H₂¹⁷O spectra from a representative voxel of 3D ¹⁷O MRSI data acquired before, during and after two consecutive 2-min ¹⁷O₂ inhalations in a rat brain at 9.4T. (B) The comparison results between two repeated CMRO₂ measurements in five rat brains. Adapted from Zhu et. al. JCBFM 2007; 27(6): 1225–1234.

3.4.4. Detectability of the CMRO₂ change

The detectability of the changes in CMRO₂ is another important aspect of the high-field ¹⁷O-MR based CMRO₂ imaging approach that requires careful evaluation for in vivo applications. It is well documented that the basal CMRO₂ is sensitive to the brain temperature (see te>|[89, 90] and the references cited therein). Most studies reported in the literature, however, were based on the global CMRO₂ measurements of entire brain using the Kety-Schmidt method te>|[74, 91] and were lacking spatial information regarding the regional oxygen consumption rate and/or its change. A CMRO₂ imaging study using a 3D in vivo ¹⁷O-CSI approach was designed and conducted at 9.4T for quantifying absolute CMRO₂ values in the rat brain at normal brain temperature (37°C) (i.e., normothermia) and mild hypothermia (32°C) conditions te>|[88]. Fig. 9A illustrates an example showing three representative slices of 3D CMRO₂ images from a rat brain under normothermic and hypothermic conditions. These images clearly show significant reduction of CMRO₂ crossing the entire brain induced by lowering brain temperature by several degrees. This metabolic suppression occurring at hypothermia was consistently observed in all five rats studied (Fig. 9B), resulting in an average of 45% CMRO₂ reduction as compared to normothermic condition te>|[88]. These findings indicate that the established in vivo ¹⁷O MR imaging approach is sufficiently sensitive for determining the dynamic CMRO₂ change and its spatial distribution resulting from physiological perturbations. Thus, the measured CMRO₂ values can be quantitatively correlated with other associated physiological parameter(s). Fig. 10 illustrates one example showing the quantitative relation between CMRO₂ and CBF: both of these were measured by the direct in vivo ¹⁷O MR approach in the α-chloralose anesthetized rat under a wide range of physiological conditions from normothermia to hypothermia te>|[88].. It clearly shows a strong correlation between CBF and CMRO₂ with a linear correlation coefficient of R = 0.97 indicating a tight vascular-metabolic coupling in the rat brain.

FIG. 9.

(A) Anatomic images (middle column) of a representative rat brain and the corresponding multi-slices CMRO₂ maps obtained using 3D ¹⁷O-CSI approach at 9.4T under normothermia (left column) and hypothermia (right column) conditions, (B) Summary of CMRO₂ results measured at normothermia and hypothermia conditions (n=5). Adapted from Zhu et. al. JCBFM 2007; 27(6): 1225–1234.

Fig. 10.

Correlation of CBF and CMRO₂ values in rat brains anesthetized with a-chloralose at brain temperature range of 32–37°C. The linear correlation coefficient (R) is 0.97. Adapted from Zhu et. al. JCBFM 2007; 27(6): 1225–1234.

4. Current status of high-field in vivo ¹⁷O MRS/MRI for studying brain bioenergetics

Since early 2000, substantial efforts have been devoted to the development of the high-field ¹⁷O MRS/MRI approach for noninvasively imaging CMRO₂ with a short ¹⁷O₂ inhalation, which were mainly carried out using small animal models te>|[12, 54, 60, 75, 79, 88, 92, 93].. These research works have demonstrated not only the feasibility but also the great promise of the high-field ¹⁷O MR approach for studying the central roles of oxidative metabolism in a living brain under various physiological conditions. Here, we highlight the major advance made in this regard, which represent the current status of the ¹⁷O MR technology for CMRO₂ imaging.

4.1. Direct imaging of CMRO₂ in animal models

For directly imaging CMRO₂ in animal models, several major steps were taken to ensure the quality and reliability of the CMRO₂ measurement. Specifically, the development of the high-field in vivo ¹⁷O-MR based CMRO₂ imaging approach has gone through the following processes: 1) feasibility assessment, where relaxation properties and detection sensitivity of the natural abundance ¹⁷O-water of brain tissue were examined at different magnetic field strengths te>|[52, 54, 92].; 2) methodology development, where a comprehensive and quantitative in vivo 3D ¹⁷O-CSI approach was established to image CMRO₂ in rat brains at 9.4T with only two minutes of ¹⁷O₂ inhalation te>|[79, 94].; 3) quantification and validation, where a complete CMRO₂ quantification model as well as a simplified non-invasive CMRO₂ imaging method were established and validated for improving reliability and reproducibility te>|[60, 75, 88]., and 4) applicability assessment, where the ability of the high-field ¹⁷O MR approach for imaging CMRO₂ and its change in animal brains under various physiological conditions were demonstrated te>|[75, 88, 93].. Accordingly, the in vivo ¹⁷O MRS/MRI approach, with a brief introduction of the ¹⁷O₂ gas, can be readily applied to image the absolute CMRO₂ in small animal models at high/ultra-high fields for various physiological, neurological or pathological studies.

4.2. CMRO₂ image of the human brain

The ultimate goal for developing the CMRO₂ imaging approach is to study the oxygen metabolism in healthy and diseased human brains. Several attempts have been made to image CMRO₂ in the human brain te>|[38, 82, 83, 95, 96].. Fiat and coworkers examined the natural abundance ¹⁷O-water signal of the human brain using 1.5T whole body clinical scanner. However, the extremely short relaxation times (T₁≤5ms and T₂≈2ms) and the poor sensitivity of the ¹⁷O-MR signal available at 1.5T led to very low spatial and/or temporal resolution of the ¹⁷O-MR imaging te>|[38, 95].. Zhu and coworkers have investigated the relaxivity and sensitivity of natural abundance ¹⁷O-water in the human occipital lobe utilizing a 7T scanner te>|[82].. The results confirmed the advantage of the high magnetic field, which had been observed in animal models, for substantially improving the ¹⁷O detection sensitivity, and which makes it possible to image the dynamic change of the H₂¹⁷O signal in human brain during a 2-min ¹⁷O₂ inhalation with excellent temporal resolution (11 sec) and reasonable spatial resolution (<1.4 cc nominal resolution). Fig. 11 displays the time courses of the H₂¹⁷O signals in two 3D-CSI voxels obtained from an representative subject before, during and after a brief ¹⁷O₂ gas inhalation (only 2 min), and it clearly demonstrates the feasibility of the ¹⁷O MR approach for imaging the human brain CMRO₂ at the high field of 7T te>|[82]. Recently, Atkinson and Thulborn applied direct ¹⁷O MR detection and ultra-short echo imaging sequence by using a 9.4T scanner for mapping CMRO₂ of entire human brain te>|[83].. The spatial (~2.7cc) and temporal (42 sec) resolutions of the 3D ¹⁷O image achieved in this study were utilized for mapping the dynamic change of the ¹⁷O water contents before, during and after ~15min ¹⁷O₂ gas inhalation in one human subject. With a three-phase CMRO₂ quantification model and co-registered ²³Na images for anatomic reference and brain tissue mass computation, 3D CMRO₂ maps of the entire human brain were obtained te>|[83] at 9.4T. The CMRO₂ values determined in this MR study were in a similar range to those reported from PET studies te>|[27]

Fig. 11.

Stacked plots of H₂¹⁷O signal measured by 3D ¹⁷O CSI at 7T before (i.e., natural abundance), during and after an 2 min ¹⁷O₂ gas inhalation from two representative voxels in the human visual cortex.

Despite these advances, quantitative and noninvasive mapping of CMRO₂ in human brain still face many technical and methodological challenges (see more discussion in Section 5). Further research efforts are needed before the ¹⁷O-MR based approach can be readily applied for studying oxygen metabolism in human brains.

4.3. Mapping functional CMRO₂ changes

One important application of the CMRO₂ imaging technique is to determine the CMRO₂ changes due to the functional activation of the brain. The BOLD (blood-oxygen-level dependence) contrast based functional MR imaging (fMRI) technique is the most widely used neuroimaging modality for studying brain function and human behavior te>|[97–100].. However, the BOLD-fMRI is unable to directly detect neuronal activity; instead, it relies on a complex interplay among CBF, cerebral blood volume (CBV), and CMRO₂ changes induced by altered brain activity te>|[101].. Precise interpretation of fMRI results requires a better understanding of the quantitative relationship between the fMRI BOLD contrast and the underlying neurophysiology—in particular the stimulus-evoked CMRO₂ change te>|[102]..

The ability to map the functional CMRO₂ changes induced by an external brain stimulus has recently been explored in a lightly anesthetized cat brain. Absolute CMRO₂ images with reasonable spatial and temporal resolutions were obtained from each cat during both resting and activation (visual stimulation) states, respectively te>|[93] and functional maps of the relative CMRO₂ changes (i.e. ΔCMRO₂/CMRO₂) were generated accordingly for each animal. Fig. 12 summarizes the findings of this study. In addition to observing significant CMRO₂ increases during the visual stimulation, the size and location of the activated brain regions depicted in the functional ΔCMRO₂/CMRO₂ maps largely coincide with those regions showing positive BOLD-fMRI changes in the same cat brain despite the different spatial resolutions of the CMRO₂- and BOLD-based functional images (Fig. 12A). In addition, by directly imaging the absolute CMRO₂ values under both resting and activated brain states (Fig. 12B), this functional ¹⁷O-CMRO₂ study not only provided a quantitative measure of relative CMRO₂ change elevated by visual stimulation (Fig. 12C) but also revealed a strong influence of baseline metabolic activity level on the relative CMRO₂ change in response to brain stimulation (Fig. 12D). Thus, this crucial finding regarding the quantitative relationship between the absolute CMRO₂ change and increased brain activity during activation suggests a tight neural-metabolic coupling and the vital role of oxygen metabolism in supporting the intensified neuronal activity in a working brain te>|[93]..

Fig. 12.

(A) Functional maps of BOLD and CMRO₂ changes during visual stimulation from a representative cat brain; (B) averaged control (baseline) and activated CMRO₂ values; (C) relative CMRO₂ changes; and (D) negative correlation between the percent CMRO₂ changes and baseline CMRO₂ values (n=6). Adapted from Zhu et. al. JCBFM 2009; 29: 10–18.

4.4. Ultra-fast CMRO₂ measurements

The ¹⁷O-MR sensitivity achievable at high magnetic field can be used to image CMRO₂ with a much improved temporal resolution while sacrificing the spatial resolution to a certain extent. An ultra-fast CMRO₂ measurement strategy has been tested in the rat model at 9.4T. Fig. 13 demonstrates an example of the dynamics of the H₂¹⁷O contents before, during and after a 2 min ¹⁷O₂ inhalation in a rat brain from measurements acquired with high temporal resolution of 1 s (see Fig. 13A); the averaged CMRO₂ value during the inhalation period in this case was found to be 1.32 μmol/g/min using the complete quantification model described in Eq. [8] (see Fig. 13B) and 1.43 μmol/g/min (linear fit) or 1.34 μmol/g/min (quadratic fit) using the simplified model described in Eqs. [9] and [10]. The simplified quantification model provides a reasonable approximation and, most importantly, it allows completely non-invasive determination of the absolute CMRO₂ value in vivo.

Fig. 13.

Ultra-fast CMRO₂ measurement in rat brain at 9.4T: (A) Dynamics of the ¹⁷O-water contents in rat brain tissue before, during and after a 2min ¹⁷O₂ gas inhalation obtained with 1 sec temporal resolution; and (B) the CMRO₂ values obtained during the inhalation period were quantified using the complete model as described by Eq.[8].

The ultra-fast temporal resolution of the CMRO₂ measurement enables the study aiming at rapid temporal changes in the oxygen consumption rate caused by instantaneous physiological or pathological alteration occurs in the animals. Two examples of such a study are shown in Fig. 14. The significant slow down or diminishing of the oxygen metabolism in the rat brain due to global ischemia (Fig. 14A) or heart arrest (Fig. 14B) is reflected in the sudden decrease or halt in the production of the ¹⁷O-labeled metabolic water te>|[103].. Thus, through careful experimental design as shown in Fig. 14, the CMRO₂ values for the two different conditions (i.e., before and after ischemia or KCl injection) can be determined with only one short ¹⁷O₂ inhalation of a few minutes te>|[103]..

Fig. 14.

Dynamics of the ¹⁷O-water contents in rat brain tissue before, during and after short ¹⁷O₂ gas inhalation obtained with ultra-fast CMRO₂ measurement approach at 9.4T: (A) forebrain ischemia and (B) KCl injection were performed during the inhalation as indicated by the arrows.

4.5. Simultaneous CMRO₂ and CBF imaging

As described earlier, the in vivo ¹⁷O MR imaging approach at high/ultrahigh field has been established for non-invasively mapping CMRO₂ in small animals. However, imaging of CBF using the same ¹⁷O MR approach usually requires invasive procedures for introducing the NMR-visible H₂¹⁷O as exogenous tracer. Experimental evidence reveals that the metabolic H₂¹⁷O water generated from a brief ¹⁷O₂ gas inhalation in the brain tissue had a much slower washout (or decay) rate compared to that of bolus H₂¹⁷O tracer, suggesting possible water permeability restrictions in the mitochondrial and/or cellular membranes te>|[75].. Nevertheless, further investigation found that the decay rate of the metabolic H₂¹⁷O after cessation of the ¹⁷O₂ gas inhalation was still closely related to the cerebral perfusion and its change; and a linear relationship between CBF and H₂¹⁷O decay rate was determined experimentally from combined CBF and CMRO₂ measurements in the rat brains under varied physiological or pathological conditions te>|[104]..

Fig. 15 shows an example of such multiple CMRO₂ and CBF measurements where the relative CBF values were also assessed using Laser Doppler Flowmeter (LDF) performed in a representative rat brain after undergoing global forebrain ischemia preparation. The metabolic H₂¹⁷O decay rates obtained in the CMRO₂ measurements during baseline, reperfusion and post-ischemia periods are displayed (see Fig. 15A) and their changes correlate well with the relative CBF changes measured by LDF (see Fig. 15B). The linear regression of the experimental data from similar measurements in different animals with different preparations for altering the brain perfusion led to the relation of CBF ≈ 1.86 × k (correlation coefficient R = 0.85), indicating that the measured metabolic ¹⁷O-water decay rate k provides a good approximation for estimating CBF in a wide range of physiological (or pathological) conditions te>|[104]..

Fig. 15.

Multiple and simultaneous CMRO₂ (A) and CBF (B) measurements using ¹⁷O-MR approach at 9.4T and Laser Doppler Flow meter (LDF), respectively, in a representative rat brain with global forebrain ischemia preparation. The H₂¹⁷O decay rates during the baseline, reperfusion and post-ischemia periods were quantified in (A) and they correlate well with the relative CBF changes as shown in (B).

The findings from these researches demonstrate that in vivo ¹⁷O MRS/MRI approach is capable of assessing not only CMRO₂ but also CBF simultaneously and noninvasively in the rat brain, and thus it provides a new utility for imaging the oxygen extraction fraction (OEF) of the brain tissue, another important physiological parameter, which is proportional to the ratio of CMRO₂ and CBF.

5. Challenge and Perspective

As described in this article, the most interesting and important application of in vivo ¹⁷O NMR is for quantitatively imaging the rate of cerebral oxygen consumption occurring in mitochondria. Questions related to this rate are encountered frequently in biomedical research when considering either normal brain function or abnormalities related to a variety of brain diseases. Relevant to normal brain function, a question of whether the alterations in CMRO₂, cerebral metabolic rate of glucose (CMR_glc) and CBF induced by neuronal activity are quantitatively coupled (or matched) is central for understanding the mechanisms underlying most modern neuroimaging techniques including fMRI and PET. On the other hand, the central role of oxidative metabolism and its metabolic rate is also evident in pathologies associated with many brain disorders. Therefore, the ability to quantitatively image CMRO₂ in vivo is essential for efforts aimed at investigating and understanding cerebral oxidative metabolism under normal and pathological conditions. The promising in vivo ¹⁷O NMR results as demonstrated during the past two decades and reviewed here have provided a crucial step towards the ultimate goal of developing a robust and completely noninvasive ¹⁷O NMR approach for imaging CMRO₂ in animal brains, and potentially in human brains.

The establishment of high-field in vivo ¹⁷O NMR for imaging CMRO₂ in small animal brains is quite successful while application of this method to the human brain faces some serious challenges. Firstly, the NMR sensitivity per unit tissue volume is reduced for human applications because the enlarged RF coil size (i.e., reduced reception sensitivity) needed for covering the entire human brain, which is approximately 700 times larger than the rat brain. This disadvantage can be partially compensated for by increasing the ¹⁷O imaging voxel size in quantification in humans and using advanced RF array coil technology. Secondly, the CMRO₂ human brain is more difficult due to the uncertainty in determining the kinetics of the ¹⁷O fractional enrichment of the ¹⁷O₂ gas and other parameters. The last but not least challenge for routine CMRO₂ imaging, especially for human study, is the cost of ¹⁷O₂ gas. Currently, the cost of the ¹⁷O-labeled oxygen gas is high because of the extremely low ¹⁷O natural abundance, low production efficiency for achieving high ¹⁷O enrichment, and presumably the low demand. Only a few companies are capable of supplying a large amount of ¹⁷O₂ now. However, it is reasonable to expect that further progress in the technical developments of in vivo ¹⁷O approaches should stimulate numerous biomedical applications including clinical diagnosis, and increase the demand, ultimately leading to more efficient ¹⁷O₂ production and lower retail prices. In addition to reduce the cost of the ¹⁷O-CMRO₂ measurement, it is also essential to improve the efficiency of the ¹⁷O₂ gas delivery system and to minimize the lost or waste of the ¹⁷O-labeled oxygen gas in the process.

In conclusion, the high/ultrahigh field NMR systems currently available or in development for both animals and humans provide great opportunities for in vivo MRI/MRS applications in medicine, especially for those nuclei with a low magnetogyric ratio. One of the nuclei that benefit the most from ultrahigh field strength is the ¹⁷O spin combined with direct in vivo ¹⁷O NMR detection, which has shown great promise for imaging CMRO₂ noninvasively. Finally, the successful developments of in vivo ¹⁷O NMR approaches will lead to an alternative or better CMRO₂ neuroimaging tool compared to the established PET method, and could have a profound impact on the study of oxidative metabolism in brains and potentially in other organs such as hearts te>|[84, 105].

Highlights.

• This article reviews the developments of in vivo ¹⁷O NMR imaging in brain research.

• In vivo ¹⁷O NMR imaging has improved significantly at high/ultrahigh field.

• In vivo ¹⁷O NMR can noninvasively image brain oxygen metabolism and perfusion.

• In vivo ¹⁷O NMR is useful for mapping the functional change in oxygen metabolism.

• In vivo ¹⁷O NMR imaging could potentially be used for human and clinic applications.

Abbreviations

γ

magnetogyric ratio

T₁

longitudinal relaxation time

T₂

transverse relaxation time

T₂*

apparent T₂

τ_c

rotational correlation time

B₀

magnetic field strength

SNR

signal-to-noise ratio

Q

RF coil quality factor

CSI

chemical shift imaging

ATP

adenosine triphosphate

ADP

adenosine diphosphate

Pi

inorganic phosphate

CBF

cerebral blood flow

LDF

laser Doppler flowmeter

BBB

brain-blood-barrier

CMR_glc

cerebral metabolic rate of glucose utilization

α

¹⁷O enrichment fraction of inhaled

CMRO₂

cerebral metabolic rate of oxygen utilization

PET

positron emission tomography

fMRI

functional magnetic resonance imaging

C_a(t)

time-dependent H₂¹⁷O concentration in excess of the natural abundance H₂¹⁷O concentration level in the arterial blood

C_b(t)

time-dependent H₂¹⁷O concentration in excess of the natural abundance H₂¹⁷O concentration level in the brain tissue

C_v(t)

time-dependent H₂¹⁷O concentration in excess of the natural abundance H₂¹⁷O concentration level in the venous blood

λ

brain/blood partition coefficient ¹⁷O₂ gas