|Home | About | Journals | Submit | Contact Us | Français|
To characterize the effect of diffusion time on short-range hyperpolarized 3He MR diffusion measurements across a wide range of emphysema severity.
3He diffusion MR imaging was performed on 19 lungs or lobes resected from 18 subjects with varying degrees of emphysema using 3 diffusion times (1.6 msec, 5 msec, and 10 msec) at constant b value. Emphysema severity was quantified as the mean apparent diffusion coefficient (ADC) and as the percentage of pixels with ADC higher than multiple thresholds from 0.30–0.55 cm2/sec (ADC index). Quantitative histology (mean linear intercept) was obtained in 10 of the lung specimens from 10 of the subjects.
The mean ADCs with diffusion times of 1.6, 5.0, and 10.0 msec were 0.46, 0.40, and 0.37 cm2/sec, respectively (P <0.0001, ANOVA). There was no relationship between the ADC magnitude and the effect of diffusion time on ADC values. Mean linear intercept correlated with ADC (r=0.91–0.94, P<0.001) and ADC index (r=0.78–0.92, P<0.01) at all diffusion times.
Decreases in ADC with longer diffusion time were unrelated to emphysema severity. The strong correlations between the ADC at all diffusion times tested and quantitative histology demonstrate that the ADC is a robust measure of emphysema.
Hyperpolarized 3He MR imaging depicts inhaled gas within the lungs. By applying diffusion-sensitizing gradients and measuring resultant changes in 3He signal intensity, the restricted diffusion of 3He within the airspaces of the lung can be quantified within each pixel as an apparent diffusion coefficient (ADC). Measuring the ADC over short diffusion times (milliseconds) has been used as a means of quantifying the increase in size of lung airspaces which defines emphysema. The 3He ADC distinguishes individuals with and without emphysema (1, 2), and is increased in many smokers with normal pulmonary function, suggesting that it is sensitive to the early stages of emphysema (3). Studies demonstrating that the ADC correlates with CT measurements of lung attenuation (4, 5) and with quantitative histologic measurements of airspace size (6–8) further support its validity for quantifying emphysema. 3He ADC measurements also appear to have good reproducibility (9–11). Consequently, the 3He ADC has been recognized as a potentially valuable biomarker for the detection of emphysema and quantification of its progression.
While ADCs obtained at short diffusion times reflect relative airspace sizes, measured ADC values are expected to decrease as the diffusion time is increased (12), due to an increase in the number of 3He atom collisions with airspace walls, reducing the mean square displacement of the atoms per unit time. For the 3He ADC to serve as a reliable biomarker for the presence and severity of emphysema, it is important to know the degree to which the ADC may change when different diffusion times are used. It is also important to know whether certain diffusion times provide ADC values that correlate more strongly with histologic measurements of airspace size or are more sensitive to disease than others. Thus, our aims in this study were to (1) test the hypothesis that ADC values decrease as the diffusion time is increased; (2) empirically characterize the effects of diffusion time on ADC measurements across a wide range of emphysema severity by determining the degree to which the ADC changes for a given change in diffusion time; and (3) determine whether certain diffusion times provide superior ADC estimates of relative emphysema severity in different subjects by assessing the correlations between the ADC obtained with different diffusion times and quantitative histology.
Nineteen resected lung specimens were prospectively obtained from 18 patients (Table 1). There were 8 lobes (7 right upper lobes, 1 left upper lobe) from 8 patients who had undergone lobectomy for lung cancer (4 female; mean ± standard deviation age 68 ± 5 years; smoking history 51±26 pack years; FEV1 88±23 percent of predicted). Nine whole lungs (6 left, 3 right) came from from 8 patients who had undergone lung transplant surgery due to chronic obstructive pulmonary disease (4 female; mean ± standard deviation age 59±6 years; smoking history 61±24 pack years; FEV1 21±5 percent of predicted). Finally, 2 whole right lungs were from 2 organ donor victims of fatal head trauma whose lungs were not matched to a transplant recipient (both male, ages 21 and 22 with limited smoking histories; no pulmonary function test results were available). The 18 patients had no clinical evidence of active lung infection, no infiltrative lung disease or more than subsegmental atelectasis on preoperative chest radiograph or CT scan, and no history of asthma or diffuse lung disease other than emphysema.
The study was approved by the local Institutional Review Board. The 16 patients who had lobectomy or transplant consented to have their resected lung specimens used in this study, while responsible family members of the 2 trauma victims consented to the use of their lungs for medical research.
As opposed to the typical in vivo use, hyperpolarized 3He MR imaging was performed ex vivo on the resected lung specimens. This enabled us to ensure distribution of the 3He throughout all of the airspaces, and it allowed repeated measurements from a single bolus of gas. Both lungs were obtained from one of the lung transplant patients and imaged separately (total 19 lung specimens from 18 subjects). All resected lobes and lungs were prepared for imaging by attaching tubing to inflate the specimen and sealing any air leaks using methods previously described (13).
The 3He gas was hyperpolarized by spin-exchange optical pumping (14) using a custom lab-built polarizer or a commercial polarizer (General Electric Healthcare) to achieve polarization levels of approximately 30–40%. To avoid the depolarizing effects of oxygen that would diminish the 3He MR signal (15), and thus to extend the utility of a single 3He bolus, the lungs were purged of oxygen with pure nitrogen gas. Immediately prior to imaging, approximately 400 ml of hyperpolarized 3He were released from a pressurized glass polarization cell into a sealed polyethylene bag via an attached polyethylene tube, and drawn into a 1 liter syringe containing nitrogen to a total gas volume of approximately 1 liter. The gas mixture was injected into the lungs or lobes to a pressure of 10–12 cm H2O, and gently withdrawn and reinjected to the same pressure two to three times to facilitate mixing throughout all of the airspaces immediately prior to starting the scan.
MR imaging was performed on a 1.5 Tesla MR scanner (Siemens Magnetom Vision) using a custom, lab-built, transmit-receive, single-turn solenoid coil tuned to the 3He resonance frequency of 48.47 MHz. Images were acquired using two interleaved 2D gradient echo pulse sequences, with a bipolar magnetic field gradient, diffusion-sensitizing pulse applied between the two imaging pulses for dephasing and rephasing of nuclear spins (1). Pulse sequences with diffusion times (duration over which 3He diffusion is encoded, defined here as the time from the start of the positive lobe to the start of the negative lobe of the bipolar gradient pulse) of 1.6 msec, 5.0 msec, and 10.0 msec were applied. The distances probed by these diffusion times approximate the dimensions of normal intraacinar airways (16, 17). The b value
where Gm is the gradient amplitude, Δ is the diffusion time, τ is the gradient ramp time, and δ is the bipolar pulse duration including ramp-up and ramp-down times (as in ref. 18, Fig. 2), was held constant at b1=0 s/cm2 for the first image and b2=1.38 s/cm2 for the second image. Other technical parameters were: 10 mm section thickness; no interslice gap; 1–5° flip angle; 64 × 64 matrix; and 35 cm field of view (5.5 mm × 5.5 mm pixel dimensions). Gradient echo time/repetition time were 6/9 msec for 1.6 msec diffusion time, 9.4/12.4 msec for 5 msec diffusion time, and 14.4/17.4 for 10 msec diffusion time. The total time to obtain one set of images from an individual specimen (20–39 sec) varied based on the diffusion time and size of the specimen.
The ADC for each voxel was calculated using custom software written in C, according to the formula ADC = ln(S1/S2)/b2, where S1 and S2 represent the signal intensities of the same voxel in the first and second images of the interleaved pulse sequences, and displayed as ADC maps of each slice (Fig. 1). The ADC for each whole specimen was quantified as the mean ADC of all pixels from all slices in each specimen having signal intensity at least 2.5 times the background noise level. In addition, analogous to the ‘density mask’ analysis method first introduced for CT quantification of emphysema (19), an ADC index was calculated as the percentage of all pixels having an ADC higher than multiple thresholds ranging from 0.30 cm2/s to 0.55 cm2/s, at 0.05 cm2/s increments. Complete analysis of a single image set (at a single diffusion time) was performed in approximately 30–60 minutes.
Quantitative histology measurements were obtained for a whole lung from 4 of the transplant patients and both of the unmatched donors, and for a lobe from 4 of the lobectomy patients. The lungs were fixed with heated formalin vapor using a method based on previously reported techniques (20, 21). In summary, each lung or lobe was ventilated under positive pressure of 12–20 cm H2O with 37% formalin vapor heated to 46°C for 4–10 hours, using a diaphragm pump with an electronically controlled circuit that provided a brief exhalation (< 1 sec) every 6–8 sec. Histologic samples for morphometry were obtained by cutting the inflation-fixed specimens into 1–2 cm-thick transverse slices (Fig. 2). Twenty tissue blocks per resected lobe and 40 per whole lung were obtained from these slices in a random manner (avoiding any tumor tissue) and embedded in paraffin. Histologic slides of 3 μm thickness were prepared and stained with hematoxylin and eosin (Fig. 2). Shrinkage of the vapor-fixed tissue after slide preparation was minimal (<5%).
Emphysema was quantified as the mean linear intercept (Lm), i.e., the average distance between alveolar walls. A single Lm value was determined for each specimen as the average of the Lm measurements from the histologic slides. To make the measurements, five random fields per slide were digitally photographed at 4X magnification. A computer program (Image Pro Plus, Media Cybernetics) was used to overlay a grid of parallel line segments on each field, automatically count the number of intersections between line segments and airspace walls, and divide the total length of the line segments by the number of intersections, to obtain the Lm for each field (8). For fields within the sample area consisting of pure airspace, the diameter of the field (1.9 mm) was used as the Lm value. The Lm values for the five fields were averaged to obtain a single value for each slide, and the Lm values for each slide were averaged to obtain the single Lm value for each specimen.
The effects of diffusion time on the mean ADC and ADC index were determined for all 19 specimens from 18 patients by performing analysis of variance using JMP 6.0 (SAS Institute, Cary, NC). The relationship between diffusion time effects and emphysema severity was assessed by Bland-Altman plots of mean vs. difference (22) using Microsoft Excel 2007. Pearson correlation was performed to assess the relationship between Lm and the mean ADC, and between Lm and each ADC index, at each diffusion time, using Microsoft Excel 2007. P values <0.05 were considered statistically significant.
The ADC values for the 19 specimens were distributed relatively evenly from 0.13 cm2/sec to 0.72 cm2/sec (Fig. 3). The mean ADC of the entire group decreased as diffusion time was lengthened, with mean ± SD values of 0.46 ± 0.16, 0.40 ± 0.16, and 0.37 ± 0.16 cm2/sec at diffusion times of 1.6, 5, and 10 msec, respectively (P<0.0001) (Fig. 1). The differences between the lowest and highest ADC values among all of the subjects were similar for the three diffusion times (0.52 – 0.54 cm2/sec), and the ADC values for the three diffusion times were highly correlated (r = 0.97–0.99). There was no relationship between the severity of emphysema as measured by the ADC and the difference in ADC obtained with any two of the diffusion times (Fig. 4). The ADC index values for the 19 specimens decreased as the diffusion time was increased at all index thresholds (Fig. 5), and were correlated to the mean ADC (r=0.93–0.99, P<0.001) at all index thresholds.
The distribution of ADC values in the 10 specimens from which quantitative histology was obtained (Fig. 6) was similar to the distribution in the entire group of 19 specimens (Fig. 3). The mean Lm for these 10 specimens was 0.55 mm, and ranged from 0.09 mm to 1.04 mm (Fig. 6). Correlation coefficients comparing the Lm with the ADC at 1.6, 5.0, and 10 msec were r=0.91, r=0.94, and r=0.94, respectively (all P<0.001).
Compared to the range of ADC values among the individuals in this cohort (roughly 3- to 4-fold), there was a greater range of ADC index values (from 12- to more than 150-fold depending on the diffusion time and index threshold). However, correlations between the Lm and the ADC index values were lower than the correlations found for the mean ADC, ranging from r = 0.78–0.87 (P<0.01) at 1.6 msec, r = 0.88–0.91 (P<0.001) at 5.0 msec, and r = 0.88–0.92 (P<0.001) at 10 msec for the different thresholds (Table 2).
This study demonstrates the degree to which the ADC for the whole lung decreases as diffusion time is increased over short time scales. Compared to a diffusion time of 1.6 msec, the ADC decreased by an average of 0.06 cm2/sec when the diffusion time was increased to 5 msec, and by 0.09 cm2/sec when the diffusion time was increased from 1.6 to 10 msec. In most published studies, diffusion times generally ranging from 1 msec to 2.3 msec have been used. Our results suggest that, although the effect of diffusion time on ADC is relatively small, it appears to be greatest in this range (see Fig. 3).
We found that the absolute change in ADC with a change in diffusion time was unrelated to emphysema severity. The average decrease in ADC of 0.06 cm2/sec when changing the diffusion time from 1.6 msec to 5 msec diffusion time was identical to the average change reported in a published abstract for three healthy subjects when the diffusion time was increased from 1.8 msec to 5.8 msec (12). However, none of our subjects showed a diffusion time-related change in ADC as large as the one subject with severe emphysema in the same abstract (from 0.67 cm2/sec at a diffusion time of 1.8 msec to 0.50 cm2/sec at a diffusion time of 5.8 msec). While we are uncertain of the reason for this different result, our measurements are consistent with predictions arising from a geometric model of the lung in which the acinar airways consist of randomly oriented branching cylinders covered by sleeves of alveoli that are open to the cylinder lumen (17).
Based on this geometric model, it has been proposed (18) that gas diffusion within each acinar airway is anisotropic at time scales in the msec range, with components DL and DT in the longitudinal and transverse directions, respectively. Since an imaging voxel always contains dozens to hundreds of acinar airways (depending on the voxel size) oriented in different directions, the measured ADC is related to these anisotropic parameters according to the equation
The dependence of DL and DT on the lung acinar airways geometric parameters was determined previously (23), based on computer Monte-Carlo simulations. These simulations showed that, in the range of physiologically relevant parameters, DL is independent of the diffusion time, and is a function only of the radial dimensions of the cylinder according to the equation
where D0 is the free diffusivity of helium in air (about 0.88 cm2/sec), r is the inner radius of the acinar airway cylinder, and R is the outer radius (inner radius plus alveolar sleeve) of the cylinder. In contrast, DT was found to be empirically related to the diffusion time according to the equation
where Ldiff = (4D0 · Δ)1/2, and was independent of r.
From equations  and , it follows that an increase in diffusion time would have no effect on DL, but would cause a decrease in DT. If the airspace enlargement in the early stages of emphysema is mostly due to an increase in r (corresponding to destruction of internal alveolar walls which limit longitudinal diffusion, and consistent with our own unpublished data), with only a small increase in the outer radius R (directly related to lung volume), then disease progression would cause a substantial increase in DL with very little effect on DT. Thus, a very similar absolute decrease in ADC (all from DT) with an increase in diffusion time would be expected at different degrees of early-stage emphysema severity, which is what we observed. On the other hand, if emphysema progression involves a large increase in R (due to acinar airway dilatation, deformation, and alveolar wall destruction), the cylindrical acinar airway geometry would be disrupted, and this theoretical analysis of anisotropic diffusion would no longer apply. In this situation of more advanced emphysema, the reason that the decrease in ADC with increased diffusion time was found here to be independent of the degree of emphysema severity is not explained by this geometric-mathematical model and requires further theoretical modeling.
A constant absolute change in ADC with different diffusion times at all levels of emphysema severity means that the percentage change in ADC is relatively greater in subjects with no or mild emphysema than in subjects with more advanced emphysema. Thus, if the ADC is to be used as a biomarker for the detection and progression of emphysema in its early stages, attention to the effect of diffusion time would be particularly important. Ideally, the same diffusion time should be maintained in cross-sectional or longitudinal studies to obtain the most precise measurements, and to make the most reliable comparisons of data from different investigations.
It should be recalled that the diffusion-attenuated MRI signal in the lungs is not a mono-exponential function of the b-value (18). Hence, the ADC determined from only two b-value measurements depends on the b-value itself. This effect, however, is small for small b-values (18), which dictated the choice of b-values used in our ADC measurements; the b-value was held constant and small for each diffusion time studied.
Because emphysema is defined by distal airspace enlargement and alveolar wall destruction, comparison of ADC measurements to quantitative histology is important in evaluating the accuracy of 3He MR imaging for quantifying emphysema. Previous studies have found good correlation between ADC measurements and quantitative histology in animal models (6, 7) and in a small population of normal and severely emphysematous human lungs (8). In our study population, which was also small but comprised lungs having a wide range of emphysema severity, the correlation between ADC and quantitative histology was also very strong. This provides further validation of the ADC as a measure of the relative mean airspace size in different lungs. The correlation was similarly strong with all three diffusion times, with no time found to be substantially superior. This is not surprising, since the ADC values with the different diffusion times were strongly correlated to each other, and the magnitude of the range of ADC values was nearly identical with each diffusion time. It also demonstrates that the ADC at short diffusion times is a robust measure of emphysema.
Quantification of emphysema by CT is based on the decrease in lung density resulting from airspace enlargement and tissue destruction. Rather than using mean lung density values, however, the most widely used CT densitometry method involves determination of the percentage of lung voxels that have attenuation values lower than a threshold level. This technique has been demonstrated to correlate well with quantitative histology (19, 24, 25). We explored the use of an analogous means of analyzing the 3He MR diffusion measurements as a potential alternative to the mean ADC, through calculation of threshold-based ADC index values. Although the range of ADC index values was much larger than the range of mean ADC values, correlations between the ADC index at multiple thresholds and quantitative histology were slightly lower than for the mean ADC and quantitative histology. Thus, our data suggest that the mean ADC is preferable to the threshold-based method for quantifying the relative severity of emphysema with 3He diffusion MR.
Because the longer diffusion times did not provide a substantially greater correlation with quantitative histology, we found no compelling reason for using a diffusion time longer than 1–2 msec when measuring short-range diffusivity. A diffusion time in the 1–2 msec range has practical benefits compared to the other diffusion times. First, there is less T2* decay for shorter diffusion times, limiting adverse effects on the signal to noise ratio. Shorter diffusion times also allow shorter scanning times for the same number of slices, or allow more slices to be obtained. Shorter scanning times in turn decrease the breath-hold period with in vivo imaging, which may reduce the risk of arterial oxyhemoglobin desaturation due to inhalation of an anoxic gas mixture, which is best avoided to optimize subject safety (26). Shorter scanning times also limit the reduction in SNR due to continued depolarization of hyperpolarized 3He atoms by the oxygen still present in the lungs at the time of the 3He inhalation. Furthermore, there is no technical reason to prefer longer diffusion times, as the current generation of MR scanners easily achieves the desired b-value at the shortest diffusion time used here (requiring the largest gradient amplitude).
There are several limitations to our study. First, the ex vivo nature of the imaging studies may affect the applicability of our results to in vivo imaging. With our ex vivo gas administration technique, we were able to insure distribution of hyperpolarized 3He throughout the lungs, even in cases of end-stage COPD. As a result, the whole lung ADC in some cases may have been greater than would be found with in vivo scanning, as severely diseased lung regions may not fill with gas during breath-holding after a single inspiration during in vivo imaging (1, 2); in theory, our average ADC should be closer to the true value. In addition, our scanning protocol involved inflating the explanted specimens to a level based on pressure rather than volume, so the imaging volume relative to the in vivo total lung capacity may have varied among the different specimens. The effect of this on ADC was likely relatively small, as it has been found that the ADC changes little with small variation in lung volume. In one in vivo study, the ADC changed by less than 0.01 cm2/sec as the inhaled volume (starting at functional residual capacity) was increased from 6% to 15% of total lung capacity (9); in an ex vivo study, doubling the lung volume increased ADC by 20–25% (27). We doubt that the lack of lung motion from cardiac pulsation in the ex vivo specimens had a measurable effect on the ADC compared to imaging in vivo. While an animal study was inconclusive on this issue (11), we have seen no variation in ADC values in the region of the heart on pixel maps of normal subjects. Indeed, with our diffusion times (much shorter than the cardiac cycle), bulk motion should generate only uniform phase shifts and cause no changes in the magnitude images employed here (1). We also note that, based on the diffusivities of ideal gases (28), our ex vivo ADC measurements at room temperature were likely minimally smaller than they would have been in vivo at normal body temperature, as such a temperature difference will reduce the free diffusivity by about 7%.
We used the Lm as the reference standard for emphysema quantification, as airspace enlargement is the defining feature of emphysema. Although our fixation method was different than the methods used by others, the Lm values we obtained were similar to those found by others (8, 25). Nevertheless, variability in the degree of inflation of the lungs after undergoing the fixation process is another potential limitation of this reference standard. We also note that the Lm may have been underestimated in lung regions with large bullae, since the measurement can be no greater than the length of the microscope field.
In conclusion, our results demonstrate the degree to which 3He ADC values are influenced by the diffusion time. The diffusion time dependence of the ADC was relatively greater for normal and mildly emphysematous lungs than for severely emphysematous lungs, suggesting that use of a constant diffusion time may be particularly important when making cross-sectional or longitudinal comparisons in the normal to mild disease range. Neither increasing the diffusion time beyond the 1–2 msec range nor the use of ADC thresholds provided any substantial benefits in terms of emphysema measurement accuracy compared to quantitative histology. Consideration of these findings may help to optimize the use of 3He ADC as a biomarker for the diagnosis and quantification of emphysema.
Supported by National Institutes of Health grants R01HL72369 and R01HL70037, and an equipment loan from GE Healthcare.
We are grateful to the patients and patient family members who allowed us to use their lungs for this study.