Inhalation of hyperpolarized 3He allows magnetic resonance imaging (MRI) of ventilated airspaces.3He hyperpolarization decays more rapidly when interacting with paramagnetic O2. We describe a method for in vivo determination of intrapulmonary O2 concentrations ([O2]) based on MRI analysis of the fate of measured amounts of inhaled hyperpolarized3He in imaged regions of the lung. Anesthetized pigs underwent controlled normoventilation in a 1.5-T MRI unit. The inspired O2 fraction was varied to achieve different end-tidal [O2] fractions ( ). With the use of a specifically designed applicator,3He (100 ml, 35–45% polarized) was administered at a predefined time within single tidal volumes. During subsequent inspiratory apnea, serial two-dimensional images of airways and lungs were acquired. At least once in each animal studied, the radio-frequency excitation used for imaging was doubled at constant . Signal intensity measurements in regions of interest of the animals' lungs (volume range, 54–294 cm3), taken at two different radio-frequency excitations, permitted calculation of [O2] in these regions of interest. The [O2] fractions in the regions of interest correlated closely with (R = 0.879;P < 0.0001). O2-sensitive3He-MRI may allow noninvasive study of regional distribution of ventilation and alveolar in the lung.
- respiratory gas
- noble gases
conventional magnetic resonance imaging (MRI) is based on signal acquisition from the magnetic resonance (MR) of spin-polarized hydrogen nuclei (protons) in water or organic molecules. Because of low proton spin density in air,1H-MRI does not visualize aerated spaces within the body. MR signal intensity, however, may be enhanced considerably, i.e., by a factor of 105, by hyperpolarization of nuclear spins beyond the naturally prevailing Boltzmann equilibrium. This can be achieved most efficiently in noble-gas isotopes with an odd number of nucleons, i.e., 129Xe or3He. It is accomplished technically by optical pumping of, for example, a3He plasma with circularly polarized infrared laser light (7). In 1994, Albert et al. (1) were the first to image the spatial distribution of ventilated lung spaces by MR tomography of inhaled hyperpolarized129Xe.
Several groups have since demonstrated that hyperpolarized noble gases, in particular 129Xe or3He (13, 15), could also become useful as imaging agents for functional MRI of the lung. In patients with pulmonary disease, 3He-MRI was used first by Kauczor et al. (10, 11). In chronic obstructive pulmonary disease and pneumonia, they demonstrated very inhomogeneous signal distribution, whereas patients with healthy lungs showed more homogeneous patterns. One obstacle to3He-MRI in patients with significant pulmonary disease and oxygenation failure was, however, that the paramagnetism of O2accelerates relaxation, i.e., the decay of nuclear spin hyperpolarization. This decay is described by a time constant of longitudinal spin relaxation,T 1. In air at atmospheric pressure,T 1 is ∼11 s, and it decreases reciprocally to O2 concentration ([O2]) (20). In effect, the presence of O2destroys the MRI signal so rapidly that functional3He-MRI studies, at least in diseased lungs, become technically demanding.
On the other hand, this problem led us to the idea that O2-induced destruction of3He hyperpolarization may be utilized for in vivo measurement of regional intrapulmonary [O2]. Until now, noninvasive techniques to measure [O2] at or beyond the subsegment level are still lacking. The objective of this study was to examine, in a large-animal model, the feasibility of using O2 effects on loss of3He polarization for the in vivo analysis of regional intrapulmonary [O2].
With institutional approval and in conformity with theGuide for the Care and Use of Laboratory Animals [DHEW Publication No. (NIH) 86–23, Revised 1985, Office of Science and Health Reports, DRR/NIH, Bethesda, MD 20892], six pigs (28 ± 1 kg) underwent anesthesia (8 mg/kg im azaperone; 1.2 mg/kg iv piritramide; 10–15 mg ⋅ kg−1 ⋅ h−1iv thiopental), endotracheal intubation facilitated by a single dose of pancuronium (4 mg iv), controlled ventilation with an O2-air mixture, and instrumentation with intra-arterial and intravenous femoral catheters for pressure monitoring and drug administration. O2 saturation of the blood was monitored by pulse oximetry.
After stabilization, the animals were transferred into a 1.5-T MRI unit (Magnetom Vision, Siemens Medical Systems, Erlangen, Germany), positioned supine, and switched to a flow-constant, volume-controlled ventilation mode on a dedicated ventilator (Servo 900 C, Siemens-Elema, Erlangen, Germany). Ventilation parameters were set at a tidal volume of 580 ml at 15 cycles/min and a ratio of inspiration to plateau to expiration of 25:10:65%. The ventilator was set up outside the magnetic field and was mounted with long extension tubing. Inspiratory and end-expiratory [O2] [fractional inspiratory O2( ) and fractional end-tidal O2( )] and CO2 concentrations were sampled from a side port at the connector of the endotracheal tube and analyzed continuously by using a commercially available sidestream respiratory gas analyzer (Capnomac Ultima, Datex, Helsinki, Finland), which performs paramagnetic O2analysis with a relative accuracy of ±2%. The experimental setup is depicted in Fig. 1.
High-degree spin polarization of3He is accomplished by optical pumping at the Institute of Physics, Johannes Gutenberg University, on campus. The gas is delivered in special low-iron-content glass vials (Schott Glaswerke, Mainz, Germany), which hold 300 ml of3He, with a polarization degree of 35–45%, at 3 bar. This container material provides sufficiently long relaxation times of hyperpolarized3He, ranging from 10 to 70 h. Loaded vials are transported within a 0.3-mT magnetic holding field to preserve spin polarization (22).
To administer hyperpolarized 3He from these vials to patients or animals, a gas control and delivery device has been developed by our group; its technical specifications are described by Lauer (12). In brief, a “bag-in-bottle” system is fed by the high-pressure 3He cell; the low-pressure bag is designed to apportion measured amounts of decompressed 3He at atmospheric pressure. When the bag within the rigid bottle is compressed from a pressurized air source, it empties, via a switch valve, its3He content at a predefined volumetric position into the inspiratory tidal volume. The within-breath range of produced by this gas-delivery apparatus has been measured, as part of the technical development of the device, at the mouth of the patient or tube connector. When used in a ventilator circuit with standard tubing during flow-constant, volume-controlled mode, and for bolus volumes between 80 and 400 ml, the applicator produces a He concentration profile that rises from 0 to 50% within 6 ± 1 ml of inspired volume and falls from 50 to 0% within 22 ± 2 ml, independent of bolus size or timing (12). All valves and switches are pneumatically driven and microprocessor controlled via a personal computer-based, self-developed application software written in Borland Delphi (version 1.0). To preserve spin polarization, the applicator is set up within the homogeneous field inside the scanner. Also, the omission of any magnetic or electronic components in its construction reduces loss of polarization during 3He administration.
Flow and pressure transducers of the3He applicator unit were calibrated with a 1-liter supersyringe; the applicator, positioned at the animal's head, was switched into the inspiratory limb of the nonrebreathing circuit next to its Y connector.
MRI of hyperpolarized 3He.
With the technique described here, MR visualization of3He amounts as small as 40 ml is possible. Depending on bolus placement within the inspiratory tidal volume, transverse supradiaphragmatic pulmonary cross-sectional, as well as coronal, images of the lungs or of the characteristics of the porcine tracheobronchial tree are obtainable (Fig.2,A-C).
MRI was performed in a clinical 1.5-T whole body MR system (Magnetom Vision, Siemens Medical Systems) with a dedicated custom-built Helmholtz transmit-receive coil (Siemens), which was tuned to the3He Larmor frequency of 48.44 MHz. The coil was centered over the lower one-third of the animal's sternum inside the magnet bore. During inspiratory apnea, nine consecutive two-dimensional 3He images of airways and lungs were acquired by using a fast low-angle shot sequence (repetition time = 11 ms; echo time = 4.2 ms; field of view = (350 mm)2; acquisition matrix, 81 × 128; image acquisition time, 1.0 s). The interscan delay was 1.2 s, with an additional delay of 0.6 s after every other acquisition. Transverse as well as (in individual cases) coronal cross sections were imaged with a slice thickness of 20 mm, with the former at a level 4 cm cephalad to the xiphoid (Fig. 2 A) and the latter in a plane including the trachea (Fig.2 B).
Image analysis was performed by using the NMRWin software [version 3.33; German Cancer Research Center (Deutsches Krebsforschungszentrum), Institute for Radiological Diagnostics and Therapy, Heidelberg, Germany]. For each individual animal and image series, large regions of interest (ROI) with a minimum size of >1,500 voxels (voxel size, 1.25 × 1.25 × 20 mm) were defined as follows (Table 1): signal intensity within a ROI had to be normally distributed, without overlap to noise distribution in a ROI of similar dimensions sampled outside the thoracic circumference. Nonventilated areas as well as major airways were excluded. In effect, this selection yielded ROIs with signal-to-noise ratios (SNR) >3. After additional correction for background noise according to Ref. 8, the arithmetic mean of signal intensity with normal distributions and SNR >3 was accepted for analysis.
Determination of [O2] by T1 measurements from serial3He images.
If, in an O2-containing atmosphere, consecutive images of hyperpolarized3He are acquired with identical MRI settings, the ratio of signal intensity (s) in a representative ROI of an image (sn +1) to that of the identical ROI in the previous image (sn), wheren is the number of image acquisition, depends on the following: 1) image acquisition per se, in particular on the flip angle α of the magnetization by a single radio-frequency (RF) pulse;2) the time interval between subsequent acquisitions, Δt =tn +1− tn ;3) the molecular density of O2( );4) relaxation properties of the tissue surfaces that come into contact with hyperpolarized3He; and5) diffusive and convective transport of hyperpolarized 3He into and out of the volume of signal acquisition.
The dynamics of the 3He hyperpolarization signal in serial lung images like those in Fig.3, A andB, can be modeled adequately, based on the theory outlined below in a simplified version, by rate equations that incorporate the above-mentioned spin relaxation mechanisms. Further details are provided in a separate MR methodology study from our group (6). With the use of measurements taken from 3He-filled phantoms and lungs, it describes the relationship between variance of signal intensities and ROI size. The effects of additional determinants of temporal signal evolution, e.g., surface relaxation and gas exchange during the imaging period, are also characterized.
The physics of NMR dictates that the reduction of longitudinal magnetization, and hence signal intensity, that occurs with any acquisition follows a function given by Equation 1wherer is the number of phase-encoding steps per image acquired and α is the flip angle in units of radian imposed by each RF excitation on the nuclear spin polarization of3He in the acquisition volume.
Simultaneously, 3He hyperpolarization and, hence, signal intensity (sn) decay exponentially, due to relaxation between subsequent images, according to Equation 2where the time constant of this decay is theT 1 of3He, which is shortened in the presence of paramagnetic molecular O2. In in vitro experiments, the following empirical relationship betweenT 1, , and temperature (T) (for T ≈ 200–400 K) in a gas mixture containing hyperpolarized 3He has been established by Saam et al. (20) Equation 3whereT 1 is in s and is a fraction of 1 amagat (ag).1At airway temperatures between 36.5 and 39.5°C, this yields Equation 3aThe combined effects of NMR excitation and longitudinal relaxation result in a signal decay (q) between subsequent acquisitions of Equation 4Incorporation of Eq. 3a into Eq.4 and neglecting residual sources of relaxation (point 4 above) as well as gas transport effects (point 5 above) yield a relationship that describes the combined effects of flip angle α and upon the decay Equation 5Flip angle α (in degrees) can be determined from Equation 6if , as the second unknown, is eliminated from Eq.6 by obtaining additional measurements at constant but at different α [or RF excitation voltage (U)]. Then the relationship between U and α is, in first approximation, linear for small α < 10° (U ∼ sin α ∼ α).
Thus for instance, two acquisition series can be performed after two separate 3He breaths, keeping and as well as tidal volume and 3He bolus volume constant but varying U by a factor of, for example, 2 (i.e., U2 = 2 U1 and, hence, α2 = 2 α1).
From the signal decays q1 and q2, which are produced by U and 2 U at the same , α (in degrees) can then be determined by simultaneous solution of Equation 7and Equation 7awhere Δt and are constants, resulting in Equation 8If, for convenience, a parameter S is introduced as a substitute for the term 1/4 (q2/q1)1/ r= S, solving for cos α yields Equation 9This flip angle α will be independent of the [O2] to the degree that remains constant between the two imaging series. With α known, it is then possible to derive from a series of signal intensity determinations, obtained in the same animal, in the same ROI and position, and with the same U, according to Equation 10In our experiments, in a ROI was calculated from signal decay of image pairs acquired at U = 10 V. Values for α were entered that had been obtained from double-acquisition series performed at U = 20 V and U = 10 V (radian α10 V ≈ 0.5 × radian α20 V), in a paired fashion at constant oxygenation conditions, slice, and ROI position.
Because were measured by sidestream respiratory gas analysis in dry gas at room temperature, the results obtained forT 1-derived intrapulmonary in a ROI (in ag at 37°C body temperature,btps) were also converted to a fractional [O2] ( ) at ambient temperature (21°C, 760 mmHg, dry) to facilitate comparison.
In this first feasibility study, the dilution of intrapulmonary gas composition by 3He and the drop in alveolar ) from the start of the imaging breath hold until the first image (2 s) were not accounted for. In contrast, the decrease of during Δt of paired imaging (1.2 s) enters into the measurement, which thus represents an average for the period of image acquisition.
Protocol and measurements.
was preset by changing the until a steady state was established. The size of the3He bolus was set to 40 ml (animal 1 only) or 100 ml (animals 2–6) by using the application unit, and the bolus was delivered after the initial 60 ml (combined volume of applicator tubing distal to the3He delivery valve and the attached inspiratory limb tubing) of an inspiratory tidal volume had been administered. Sequential MR images were acquired during breath hold after end-inspiration had been reached. Before and after each apneic episode, was recorded, together with end-tidal CO2 fraction, expiratory minute volume, and vital parameters.
Because, in this specific experimental setting, measurements of functional residual capacity (FRC) and serial dead space were not available, we approximated the reduction of preapnoic , which occurs by dilution with 100 ml of3He theoretically, using , expired tidal volume, and estimates for dead space volume (200 ml) and FRC [30 ml/kg (16)]. Under these conditions,3He bolus administration was estimated to reduce of a subsequent normal expiration, on average, by ∼8%.
Study interventions consisted of variations of the RF excitation and of : for flip angle determination, consecutive series with RF excitations of U = 20 V and U = 10 V were performed; pilot studies had shown that an U of 20 V was the prevailing source of signal loss at a known steady-state . The experiments designed to determine regional at several levels of used the smaller RF excitation of U = 10 V; this lower voltage was chosen such as to permit O2 to exert more influence on signal decay in relation to RF excitation. In the same ROIs, signal intensity ratio q10 V = s2/s1was then determined from paired images obtained at 2 s and 3.2 s after inspiratory hold, giving a Δt of 1.2 s.
Applicator settings for 100-ml boluses of3He in animals 2–6 resulted in the delivery of 100 ± 12 ml (means ± 1 SD for a total of 41 bolus administrations). For reasons of comparability, only those image series (n = 29) that fulfilled the following conditions were included in further analyses of flip angle and O2 density: a measured bolus size between 90 and 110 ml of 3He, first image acquired 2 s after inspiratory hold, and resultant ROIs with SNRs >3. Thus 12 paired series, performed with U = 20 V and U = 10 V at the same , could be used for calculation of flip angles in the defined ROIs. Seventeen 10-V image series were useful for determinations; the minimum SNRs of ROIs chosen from these images ranged between 3.2 and 10.8.
Linear regression analysis was used to describe the strength of the relationship between the measured with the respiratory gas analyzer and instantaneous fractional [O2] ( ) as determined by the above method in two-dimensional3He images of the lungs.
Qualitative evaluation of image series.
In Fig. 2 A, a representative two-dimensional 3He image of an animal's lungs is shown, acquired, at a level 4 cm cephalad to the xiphoid, in end-inspiratory hold after administration of a 95-ml bolus of hyperpolarized 3He within the initial one-half of a tidal volume. The effects of different intrapulmonary and doubling of RF voltage were already discernible in preliminary image series, which were obtained in the first two animals (animals 1 and2) during inspiratory breath holds after steady-state had been established and 3He volumes of only 40 ml (animal 1, Fig.3 A) or 100 ml (animal 2, Fig.3 B) had been administered. Figure3 A illustrates that O2 in the lungs accelerates3He signal decay in a dose-dependent fashion. Each four-image series taken at a different represents a period of 6 s of inspiratory breath hold. Because the applied 3He amount in thisanimal 1 (40 ml) differed considerably from that of the next series, its image data were not included in the subsequent quantitative O2 study. Figure 3 B demonstrates, in images taken 3 s apart, that the signal of a 100-ml bolus given at similar levels of is also destroyed much quicker when the RF voltage is doubled. Paired-image series like this one were used to extract flip angle α, as described in methods and below. Sizes of ROIs that provided acceptable signal quality are shown exemplarily in Fig. 4,A andB. Table 1 gives an overview of their number and their sizes in relation to the total cross-sectional lung field.
Determination of flip angle α in ROIs of3He representations.
In each of animals 2–6, identical ROIs of 10- and 20-V image pairs acquired at the same were compared to extract the influence of RF excitation on signal intensity. By using Eq. 9 , flip angles α were determined for each ROI from the ratio q2/q1of signal intensity decays q2 at U = 20 V and q1 at U = 10 V. α10 V, as the flip angle that corresponded to RF excitations with U = 10 V was found to be 3.11 ± 0.14 (SD)° (n = 12 pairs and ROIs). Individual, i.e., ROI-specific, results for α10 V are listed in Table1. α10 V showed no correlation to the at which they were determined (R 2 = 0.001).
3He-MRI-based determination of Froi .
Instantaneous fractional [O2] during the third second of an inspiratory breath hold, determined from extrabronchial ROIs of two-dimensional 3He images of the lungs ), correlated with the measured immediately preceding the apnea (R = 0.879;P < 0.0001). Figure5 demonstrates the data points of individual animals.
The y-intercept of the regression line, as an extrapolation toward a of zero, was not significantly different from zero ( at = 0, +0.027; P = 0.495). This suggests that any systematic contribution of factors other than O2 and α, e.g., wall relaxation, to signal destruction was very small.
The slope of the regression line of 0.90, on the other hand, indicates that , which had been determined after one3He-containing breath, appeared systematically lower than the measured during steady-state3He-free ventilation (Fig. 5). The reduction of alveolar gas concentrations by3He admixture, estimated inmethods to be ∼8%, would appear in the regression as an increase of the slope to 0.98 without changing they-intercept or the correlation coefficient.
This study confirms the O2dependence of longitudinal nuclear spin relaxation timeT 1 of3He in vivo and shows that this effect can be used to determine instantaneous, regional [O2] within the respiratory tract. The results are based on the feasibility of administering measured amounts of hyperpolarized3He reproducibly within single inspirations; this allowed us to follow quantitatively, by MRI, the temporal evolution of 3He hyperpolarization within the lungs.
Application of hyperpolarized 3He in MRI.
Hyperpolarized 129Xe or3He gas can be visualized by MR techniques. Static imaging of the bronchopulmonary airspace has already been demonstrated in animals (1, 15), human volunteers, and patients (10). At relatively low grades of hyperpolarization (5–20%),3He has also been administered during controlled ventilation to be visualized by repetitive ventilator cycle-gated image acquisition (3). This was performed during 10–40 inspirations to increase overall signal intensity and to compensate for O2-dependent signal attrition in vivo.
We use, instead of the common Rb exchange polarization (4), so-called metastability exchange optical pumping of3He (5), which is highly efficient in terms of rate, as well as grade, of hyperpolarization. Meanwhile, 35–45% hyperpolarized 3He is achieved routinely (7). We have developed special storage and transport techniques, which preserve longitudinal relaxation timeT 1 up to 120 h (9). High-grade hyperpolarization allows the imaging of single inspired3He volumes with high SNR (2, 7,10, 11).
The relaxation time constantT 1 of hyperpolarized 3He due to O2, which has been quantified in vitro by Saam et al. (20), will recover on an average of ∼14 s within the lungs of volunteers breathing air to up to 36 s if two to four large tidal volumes of N2 or4He are inhaled to wash out O2 before3He inhalation (2). For clinical purposes, particularly in sick patients, however, anoxic flushing of the lungs followed by prolonged apnea is not feasible.
A prerequisite for patient studies was, therefore, to develop a method of reproducible tracer-gas administration. It should allow a timed, quantitative 3He delivery into any portion of a vital capacity inspiration, be independent of the mode of respiration or ventilatory support, and cause minimal loss of hyperpolarization during administration, and O2 effects should be quantifiable. This has been accomplished with sufficient accuracy and precision by our prototype computer-controlled, nonrelaxing gas control and delivery device.
Thus this technique of exactly timed, flow-controlled administration of high-grade hyperpolarized 3He, as it is described here, is a new addition to functional MR imaging per se. Despite the use of comparably small3He amounts, a good SNR can be achieved from the upper airways to the alveolar space. This technique may better accommodate patients with pulmonary disease and may be less prone to motion artifacts than prolonged repetitive imaging, as required by the method of Black et al. (3). For instance, very fast imaging sequences may eventually allow the clinician to analyze the distribution of ventilation in patients with lung disease without radiation exposure, at a temporal resolution in the range of 100–150 ms/image and a spatial resolution of 2–3 mm.
Intrapulmonary O2 measurement.
O2 confounds morphological interpretation of MR images of airspaces, filled with partially hyperpolarized 3He. In a first step, we had designed a device for automated3He application to reduce random O2 effects, because3He volume, hyperpolarization grade, time of exposure to relaxing environments, and placement in the respiratory tract could be kept fairly reproducible. By producing the highest grade of hyperpolarization obtainable and preserving it as well as possible before administration, our goal was to reduce the inspired amount of indicator gas to minimize its disturbance of the observed system. The applicator device, together with high-grade3He hyperpolarization and appropriate MRI sequences, enabled us to use the O2-T 1relationship for [O2] measurements in vivo.
Because it was decided to use MR and respirator equipment already certified for human studies, we chose a large-animal model that could be accommodated by a similar setup and, in particular, the transmit-receive coil. Adolescent pigs were selected for several reasons. Anatomic dimensions and respiratory characteristics are closer to human conditions than those of small animals. Distribution of ventilation is well studied in pigs, its temporal heterogeneity is known to be small (18), and there is lack of collateral ventilation (19).
Numerous technical and physiological factors may be responsible for the residual variance of , which is not explained by in our study. In this first series, we compared respiratory gas samples expired from the entire alveolar space with the MRI data derived from one imaged coronal slice. At this stage, we did not aim for maximum resolution but sought to demonstrate, in principle, the feasibility of image-based O2 measurement within the lungs. Comparability of the data sets from the two measurement techniques is limited, first and foremost, by our present two-dimensional imaging technique, which provides only one slice for analysis. The required large SNR restricted analysis further, i.e., to ROI instead of total lung cross sections. We, therefore, selected rather large ROI with normally distributed signal intensity histograms, which allowed us then to average signal intensities of all pixels within these ROI and to determine a mean . Because high SNRs will represent predominantly areas with good local ventilation (high 3He content), a selection bias toward well-ventilated regions may be expected; possibly, it is reflected in the residual 2–3% overestimation of by our MR-derived mean . On the other hand, exclusion of low-signal regions gives also some preference to zones containing low , where signal intensity in the second image of a pair is preserved somewhat better. The net effect of this selection bias remains unclear at present and may differ depending on .
Mixing of the steady-state that existed in a ROI before the 3He breath with the inspired 3He bolus introduces systematic error that is unavoidable with our present application technique. The size of this systematic error depends on the relation of the indicator gas volume to its intrapulmonary volume of distribution. From our results, we estimate it to be 8–10% in our study.
Within a single inspiration, regional distribution of the inspired air-O2 mixture, as well as an interposed indicator gas bolus, should all be governed by the regional distribution of flow resistances. The within-breath range of resulting in peripheral airways from the bolus technique of3He administration, however, has not been determined with our system, but we assume that the relative admixture of 3He into regional alveolar gas does not vary so widely as to explain the entire variance observed. Indicator gas and respiratory gas distribution and mixing may differ, however, if transitions to turbulent flow occur in different regions for 3He and the air-O2 mixture. Such variation may increase with regional inhomogeneity of ventilation and may explain part of the variance of found in our series, which is much larger than that possible from ventilation-perfusion (V˙a/Q˙) heterogeneity alone. On the other hand, in healthy supine beagle dogs ventilated with He-O2 or air or SF6-O2, gas-composition-related flow characteristics have not been found to be major determinants of intrapulmonary gas transport (21), and diffusive gas transport appeared to overwhelm small differences in convective gas transport. Also, in our study, images were obtained during breath holds of ≥1 s. In pilot experiments, we have performed ultrafast MR imaging of an 80-ml 3He bolus followed by air tidal volumes into a lung phantom (500-ml silicone bag); at the end of a 0.8-s inspiration, MR signal distribution within the bag was already homogeneous (unpublished data). This suggests that gas mixing in the alveolar space may have also been complete during imaging.
At any rate, because of its lower mass density and higher diffusivity, He will be the least affected of all indicator gases by regional differences in the distribution of ventilation. Theoretically, dilution of “true” could be prevented by ventilation to a steady state of gas exchange with a mixture of He, O2, and air and then performing inspirations containing hyperpolarized3He with unchanged . Rapid relaxation of hyperpolarization during the time spent in an O2-containing mixing chamber makes this approach technically difficult, and ventilation with a He-containing gas mixture would also constitute some deviation from normal physiological conditions.
The decrease in during the imaging breath hold also enters the measurements, as outlined inmethods. Thus variability in instantaneous O2 uptake among imaging series as well as among the ROIs may also have influenced our results, although this source of error may be minor in anesthetized animals during breath-hold periods of only 2–4 s. It illustrates, however, that instantaneous at the beginning of an inspiratory breath hold, even if measured with perfect accuracy and precision, will be different from in a lung region or compartment, as determined by techniques measuring in a steady state of ventilation and perfusion (e.g., multiple inert-gas-elimination technique).
Several other determinants of regionalV˙a/Q˙ ratio, e.g., topography, anteroposterior gradients of perfusion and ventilation, and oxygenation itself, may have varied with each acquisition and region. For instance, the supine position used in this series may have slightly reduced the uniformity ofV˙a/Q˙ distribution; on the other hand, the two-dimensional, coronally oriented ROI contained both ventral and dorsal areas and their respective, more or less uniform contributions to measured end-expiratory O2. Although we studied healthy animals, additional influences known to increaseV˙a/Q˙ mismatch, like anesthesia, positive-pressure ventilation, and neuromuscular paralysis, were also present in our experiment. Under these circumstances, is never devoid of contributions from under- or nonperfused alveoli and will, therefore, fail to reflect “ideal” (17) as well as of an arbitrarily selected ROI.
It was thus to be expected that , when determined in ROIs of variable size and location, varies intra- and interindividually but should maintain a fair overall correlation to global . Our data from these experiments confirm, however, that [O2] in lung regions of a large animal model can, in principle, be determined from the decay of 3He hyperpolarization. Variability of regional intrapulmonary [O2] due to regional heterogeneity in pulmonary ventilation and perfusion may well contribute to the rather large variance observed, but one would expect its range to be smaller than that seen in our experiment. In resting healthy subjects, in whom contamination by lowV˙a/Q˙ alveoli or gas from alveolar dead space is likely to be minimal, end-expiratory can be assumed to approximate mean alveolar values, whereas a widened scatter ofV˙a/Q˙ ratios, and hence local (16), is characteristic of pathological conditions like chronic obstructive lung disease, emphysema, or asthma (23). To date, there is no reference laboratory method or clinically applicable technique to measure true regionally and to resolve its scatter. Further development of the method described in this paper into a technique of high-resolution mapping might have the potential for application to a broad range of human pulmonary diseases.
At the present stage, however, interpretation of our results still has to take several physical and technical limitations into account. The in-plane spatial resolution physically attainable with our present imaging system is in the order of 1–2 mm2. The minimum ROI size for O2 determination, on the other hand, depends critically on the SNR obtained. Sufficient3He entry, i.e., at least a minimum of regional ventilation, is of course a prerequisite. SNR then improves for a given voxel size with the degree of3He hyperpolarization, i.e., with higher spin density. High hyperpolarization allows the reduction of the amount of inhaled 3He, thus improving the accuracy of determination, and the voxel size, thus improving spatial resolution. Presently, 50% hyperpolarization is attainable on a regular basis with our system.
Loss of hyperpolarization by RF excitation and O2 relaxation will be accompanied by simultaneous signal intensity changes of other origin: diffusion of He into and out of the volume of acquisition will occur but may be restricted in peripheral lung parenchyma because of the absence of large conductive airways and also, in the porcine species, of collateral ventilation (19). Resolution will be affected further by both [O2] and3He diffusion, because effectiveT 1 depends on all local [O2], which are encountered by hyperpolarized 3He atoms during the observation period. Relaxation by bronchoalveolar surface contact, on the other hand, has not appeared as a significant contributor to T 1reduction, suggested by the very smally-intercept in our correlation (Fig.5), as well as by findings in O2-depleted pig lungs (6).
Inspired 3He dilutes alveolar [O2] depending on the3He amount, distribution of ventilation, and lung volumes. The latter are known to change during anesthesia and artificial ventilation and during inspiratory hold. Because FRC was not measured, reduction of by dilution with the tracer gas was estimated only. The numerical effect of this correction on the slope of regression lines in Fig. 5suggests that dilutional effects are likely and that the measured before breath hold overestimated the3He-diluted , which would have been expired at the time of imaging. Ongoing O2 uptake and inert-gas concentration due to alveolo-capillary O2 transfer during the 2–4 s of apnea may have added to this bias. In the phantom experiment mentioned above, MR analysis of the air-3He mixture in the respirator bag yielded a of 0.195 bar, with a Po 2 of 0.193 bar determined by gas analyzer (6).
Another source of artifact, suggested by observations during fast-inspiratory imaging sequences, is inadvertent replenishment of hyperpolarized 3He from outside the target volume during acquisition. This may occur either by3He convection along airways or by movement of 3He-containing lung tissue into the imaged slice with cardiac oscillation or with relaxation of compressed respirator tubing. Particularly during imaging of thin two-dimensional partitions, convective and diffusive movement of gas or the lung itself into and out of the imaged slice, despite the inspiratory hold, appears as an important source of error. In two-dimensional studies as this, this effect may limit the reduction of slice thickness and, hence, spatial resolution. This effect has also been studied by our group (6). Multislice two-dimensional imaging or three-dimensional imaging with simultaneous excitation of all spins may solve this problem in the future and may permit regional O2 determinations in lung volumes of 1–2 cm3.
Finally, some signal intensity variation from RF coil inhomogeneity constitutes another possible source of error. To eliminate these shortcomings, a detailed MR physical error analysis is presently under way, an improved transmit-receive coil has already been constructed and tested, and appropriate three-dimensional imaging sequences are being developed.
In summary and perspective, our study shows that3He-MRI holds promise as a functional imaging technique to analyze the regional distribution not only of ventilation but also of [O2] within the lungs. Its potential spatial and temporal resolution in the range of cubic millimeters and seconds, its noninvasiveness, and the almost unrestricted repeatability of the measurements appear as possible advantages over present techniques based on radiography, radioisotopes, microspheres, or aerosols. The method may eventually contribute to the resolution of O2 kinetics within functional lung units and help to identify regions ofV˙a/Q˙ mismatch in diseased lungs.
This work was funded by Deutsche Forschungsgemeinschaft Grant TH 315/8–1; by Innovationsstiftung Rheinland-Pfalz, Mainz, Germany; and by Institut für Diagnostikforschung, Berlin, Germany.
Address for reprint requests and other correspondence: B. Eberle, Dept. of Anesthesiology, Johannes Gutenberg Univ., School of Medicine, Langenbeckstr. 1, D-55131 Mainz, Germany (E-mail:).
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.
↵1 One amagat (1 ag) = gas density/(2.68675 × 1019molecules/cm3).
- Copyright © 1999 the American Physiological Society