Background
X-ray computed tomography (CT) is excellently suited to visualisation and accurate measurement of air volume in the airway tree [
1]. CT scanning provides maps of the linear attenuation coefficient of x-rays at an average energy of 70 kV. At this energy the attenuation coefficients of low atomic number tissues such as soft tissue are closely proportional to density. Thus CT images of the lung reflect density and therefore have the advantage of providing accurate values of the air volume in the lung. They give images of the regional distribution of air volume in addition to enabling values of total lung volume to be assessed. With multi-detector high resolution CT, the whole lung field can be imaged in less than 10 s enabling images to be obtained while subjects hold their breath at a particular level of respiration. Images at total lung capacity are most common, but images at other levels of respiration can be acquired [
2]. Image analysis of the CT data allows segmentation of approximately the first six generations of the airway tree, enabling accurate definition of the geometry of this section of the airway [
3]. This information can be applied to derive functional parameters using computational fluid dynamics [
4]. While the peripheral and alveolated airways are too small to be visualised individually, the pattern of their appearance in the images can be used to characterise typical patterns occurring in different diseases. A simple example is the assessment of the density of the peripheral area to give an emphysema score [
5]. More complex pattern recognition algorithms can be used to identify the appearance of different diseases [
6]. In particular, a method involving parametric response maps derived from CT data acquired at full inspiration and full expiration has been described, which is useful in distinguishing between emphysema and functional small airways disease [
7].
CT-derived parameters have also been used to follow developmental and compensatory lung growth [
8] and to assess structural changes in idiopathic pulmonary fibrosis [
9] and emphysema [
10].
Another area where measurements of regional lung volume are used is in normalisation of measurements of the regional distribution of aerosol deposition using radionuclide imaging. The relative deposition in conducting and peripheral airways is often assessed by a central to peripheral ratio. The values obtained depend quite significantly on the size of the areas chosen. This dependency can be considerably reduced by normalising the central to peripheral ratios to lung volume [
11,
12]. Lung volume is often estimated using transmission scanning or ventilation imaging. Both methods have disadvantages – delineation of the lungs on transmission is imprecise and ventilation does not reflect lung volume in poorly ventilated areas. The ability to use CT for this purpose overcomes these limitations. The disadvantages of CT are its relatively high radiation dose and its accessibility. Dose may be reduced using low dose protocols [
13] and the use of CT in conjunction with both gamma camera tomography (Single Photon Emission Computed Tomography, SPECT) and positron emission tomography (PET) is becoming increasingly commonplace.
Measurements of lung anatomy from CT have a further application in the field of inhaled aerosol deposition. They provide anatomical information which allows improved computer modelling of the deposition pattern. In this respect images of lung volume obtained at mid-tidal breathing such as those obtained in some SPECT-CT protocols are particularly useful in that they provide information on lung anatomy at the same stage of breathing as is frequently used in experiments on aerosol inhalation.
An important disadvantage of CT, and indeed a fundamental problem of most three-dimensional imaging studies, is that the images are acquired in the horizontal position. It is known that lung anatomy varies very significantly between supine and erect positions; the FRC of the lung is greater by around 40% with the subject erect compared to supine [
13]. Therefore all information regarding lung anatomy including lung volume values have to be compared in the light of this large effect.
In addition, it has been shown that even when volume measurements in the supine position are compared, there are significant differences between values from CT and other methods, with CT values being generally lower than other techniques [
14‐
16]. While CT scanning is now mentioned in standard texts on lung volume assessment [
1], there is a shortage of reference values for expected lung volume values in control subjects. Reproducibility data is available for CT lung volumes obtained at total lung capacity and residual volume [
17]. Also, some work has been done on reproducibility of lung volume values from CT in preclinical studies [
13].
In this study measurements of lung volume have been made using low resolution (and therefore low dose) CT during tidal breathing in 11 healthy subjects in the supine position. This has enabled mean values of total lung volume and patterns of regional distribution to be obtained at a mid-tidal breathing position, together with measures of inter-subject variability. Each subject was imaged on two separate occasions enabling the intra-subject reproducibility of the measurements to be assessed. The results presented in this paper are derived from data acquired in a previously described experiment [
18]. The prime purpose of that work was to provide data for validation of computer models of aerosol deposition by comparison with experimental measurements using medical imaging. The information on lung volume described herein is a by-product derived from that image dataset.
Discussion
This study has demonstrated that it is possible to use low resolution, and therefore low dose CT, to provide quantitative images of lung volume at approximately mid-tidal breathing. These depict the fractional lung air volume per unit space volume and they enable, among other things, the total air volume of the lungs to be obtained. A normal range for this has been defined in a set of 11 subjects and an assessment of intra-subject variability between repeat measurements has been obtained. A method for producing a mean lung volume map has been described and used to provide such a map for this group of subjects. This has been analysed to produce a description of the regional variation of fractional air volume from (i) centre to periphery of the lung using a shell analysis and (ii) as air volume profiles in three orthogonal directions. The reproducibility of the central to peripheral ratio of air volume derived from the shell analysis has also been measured.
The methodology described in this paper is not fully automated, and therefore not ideal for routine use. However algorithms are now commercially available, e.g. from VIDA Diagnostics, USA, which are capable of fully automatic segmentation of the lung and determination of lung volume. It will be interesting to see whether this software is capable of being used with low dose images such as those described in this paper and also whether the values obtained of lung volume at mid tidal breathing are similar. A key aspect of the segmentation of the lung and the subsequent analysis is the definition of the position of hilum of the lung and how much of the area around the hilum is included in the lung envelope. In this paper we have used the first bifurcation of the main bronchus to define the hilum and have used a technique involving dilation and erosion to fill in the areas in the lung envelope with high density including those around the hilum. It would be useful for the lung imaging community to make an effort to standardise these definitions so that results of different software will be comparable.
The results on total lung volume are now considered. As expected, the right lung space volume was found to be significantly higher than the left lung, contributing 53.9% of the total space volume of the lungs. The average fractional air volume concentration was also significantly higher for the right lung with the result that the fraction of air volume from the right lung is slightly but significantly higher at 54.8%. Consistent with these findings, the averaged coefficient of inter-subject variability was found to be higher for the air volume than for the space volume, 32% compared to 27%. The inter-subject variability of fractional air volume concentration was considerably lower at 8.6%.
The average fractional air volume concentration in each lung correlated significantly with the total space volume of the lung. This is in agreement with the study by Gevenois et al. [
25], who observed the fractional air volume at full inspiration in the supine position correlated with total lung capacity in healthy volunteers. This correlation may be explained by considering the gas exchange requirements of the lung. If density were to remain the same as body size increases then both the space volume and the air volume of the lung would increase proportionately. Assuming that alveolar volume is the largest component of lung volume, the volume of alveoli will then increase in proportion to body size. However, the surface area available for gas exchange will only increase proportional to body size to the power two thirds. Therefore to allow for this, the alveolar volume will have to increase at a rate which is greater than proportional to body size to maintain the required increase in alveolar surface area. It should also be noted that the difference in FAVC between left and right lungs also followed this pattern with the larger right lung also having the higher FAVC. This variation will be important to bear in bind when interpreting results of density measurements in the assessment of emphysema. Since most diagnostic CT scanning is carried out at full inspiration or full exhalation with high resolution CT, it will be important to confirm this finding in subjects imaged at these states of inhalation.
The mean total, (i.e. left plus right), mid-tidal air volume in the supine position used for imaging was 1900 ml. This is reduced by 43% from the erect FRC value obtained in these subjects using helium dilution of 3332 ml. Using the standard equation described by Stock and Quanjer [
26], the mean predicted erect FRC for these subjects was found to be 3377 ml. This is very close to the mean value found, so this group of subjects can be considered to represent a typical adult male population. Ibanez and Raurich [
27], found the average male supine FRC using He dilution to be 2230 ml. Therefore the CT values found in this paper are reduced by 330 ml compared to the helium dilution technique. One reason why the mean value obtained from CT will be lower is that it only describes the air volume from the lung hilum onwards. It does not include the nasopharynx, oropharynx, trachea and main bronchi. All this volume will be included in a helium dilution measurement. The mean total volume of this space in adult males is 226 ml. The components of this are nasopharynx, 138 ml [
26], oropharynx, 61 ml [
28], trachea, 19 ml [
29] and main bronchi 8 ml [
30]. This can go some way to explain the lower value obtained with CT. However, this will be balanced by the fact that a CT image obtained during tidal breathing might be expected to be higher than FRC as it should be equal to FRC plus half the tidal volume. Since the mean tidal volume is around 500 ml this means that CT values might be expected to be around 250 ml higher. This almost exactly balances out the underestimation described above due to the missing extra-pulmonary volume.
The above arguments lead us to consider whether there is any technical reason for an underestimation in the CT measurements. The most likely explanation here is due to the partial volume effect resulting from the limited spatial resolution of the scans. A first order correction for this effect has been applied as described above. However this only uses a global correction to the missing volume close to the surface of the lung and it may not apply accurately to all parts of the lung where local effects may be important. For example towards the base of the lung, the organ becomes very narrow and in some slices there is a thin section of lung volume, where no voxel may come above the threshold. The global correction applied will allow for this to some extent, but it may not be sufficient. This effect will be greater in low resolution CT studies such as those described here, particularly as mean tidal breathing has been used as compared to a breath hold. Investigation of the magnitude of this effect using image simulation techniques will be the subject of further study.
The correlation of CT lung air volume with FRC was reasonably good (COV 15%) and comparable with previously described results [
31]. CT imaging might be considered as an alternative method for measuring FRC, although the large difference in supine compared to erect values would need to be taken into consideration in interpreting the results. Further studies comparing CT and helium dilution should ideally measure a supine FRC with helium dilution to enable a direct comparison of results. The inter-subject variability of CT lung air volume (32%) was not significantly different from that of FRC (20%). However it was considerably larger than for FRC. Increased number of subjects will enable clarification of whether this difference is significant.
There was good intra-subject reproducibility for total lung air volume in these control subjects, with a coefficient of variation of repeat measurements on separate occasions being 6.5%. This compares favourably with previous assessments of reproducibility [
13]. The intra-subject coefficient of variation of average fractional air volume concentration was even lower at 3.1%. However this reduced variation might be expected given the lower inter-subject variation for this parameter compared to total air volume.
CT scans of the lung give the opportunity of investigating the three dimensional distribution of air volume in the lung at various stages of respiration. Most work to date has centred on looking at the number of voxels with very high air volume concentration at full inspiration as a measure of emphysema [
5]. The values obtained for this parameter will be dependent on the precise imaging protocol [
32] and analysis of these low resolution scans of the airway tree at average tidal breathing are not likely to be useful for this purpose. The elastic registration of each image to a standard template has enabled a mean image of the distribution of air volume to be obtained. This provides a database of the normal distribution of air volume against which air volume distributions of subjects with suspected lung diseases might be compared using techniques such as statistical parametric mapping [
33]. The method of non-rigid registration described in this paper uses a simple global spatial transformation, suitable for the relatively low resolution spatial images acquired. More sophisticated techniques would be required for high resolution images where fiducial markers such as the bifurcation points can be identified using commercially available software (e.g. Apollo, VIDA Diagnostics, USA). These would allow a more precise local deformation to be applied.
Two methods of analysing the spatial distribution of air volume in terms of one-dimensional vectors have been described in this paper. The first is the radial transform or shell analysis technique which allows a description of the variation of air volume concentration from centre to periphery of lungs. The different shells defined in the analysis correlate approximately with airway generation number and therefore provide a first order approximation to distribution of air volume concentration with airway generation. This form of analysis has been used in describing the three dimensional spatial variation of inhaled aerosol deposition from the centre to the periphery of the lung [
24]. In both two dimensional [
11] and three dimensional [
34] imaging of aerosol deposition, it has been found valuable to normalise the shell data to the distribution of either ventilation or lung volume. Normalisation to ventilation has the disadvantage that it may not be valid in diseases of the lung with abnormal ventilation. Normalisation to lung air volume seems a more robust method and, as CT provides the best imaging of lung air volume distribution available, it could be regarded as the gold standard method for this type of normalisation. The finding that the central to peripheral ratio of lung air volume was different for the left and right lungs shows that normalisation of aerosol deposition should be carried out separately for each lung. The intra-subject reproducibility for the c/p ratio of 8.5% was considered to be reasonably good, although more variation was found for this parameter compared to measures of the total air volume in the lung.
The other method of analysis described is the one dimensional analysis of fractional air volume concentration in three orthogonal directions. This has provided some interesting observations on the trends of this parameter in three dimensions in the supine position. The data provides a baseline against which other measurements of these profiles can be compared and it will be interesting to observe how consistently the variation described in this paper is found in future studies.
There were also differences in the distribution of air volume between the left and right lungs. The one dimensional profiles illustrated the lower values of FAVC for the left lung but the variations in the three orthogonal directions all showed similar patterns of variation for both lungs. The variation of FAVC with shell number again showed the consistently lower value for the left lung, but the relative difference was much greater for the inner shells resulting in a significantly lower central to peripheral ratio for the left lung. This is presumably as a result of the relatively higher proportion of major vessels in the reduced space volume of the lung envelope on the left side, due to the presence of the heart.
The results described in this paper on lung volume in the mid-tidal position are a by-product of a study carried out for a different purpose. For this reason not all aspects of the experimental design are optimised for lung volume measurement. Nevertheless, it is considered that useful novel results on the use of CT scanning to measure total and regional lung volume have been obtained. The number of subjects studied for example was not particularly optimised for this study. However from the repeat measurements on the 11 subjects studied, reasonably precise measures of average parameters of lung volume and repeatability have been calculated. In addition statistically significant conclusions on differences between parameters and correlations have been made. Comparison of the mid-tidal volume measurements obtained in the supine position with erect measurement of FRC was not ideal. In principle FRC would have ideally been carried out supine at the same time as the CT scan was being acquired, so that direct comparison of the measurements could be made.
Further work required on the spatial analysis of the distribution of volume is a description of the fractional air volume concentration by lobe and by segment. Since it is now possible to delineate lobes directly from CT images, and estimate the outline of the segments, lung volume parameters can be expressed per lobe and per segment. Initial work has shown that airway volume does vary regionally [
35] and this will be the subject of future study.
The determination of regional lung volume at a particular point in the respiratory cycle using CT gives the possibility of looking at change in lung volume between different points of the cycle. The recent development of four dimensional CT (4DCT) enables dynamic CT imaging to be carried out over the respiratory cycle [
36]. This would enable visualization of regional changes in volume, leading to high resolution quantitative ventilation imaging. This could potentially give important new information on lung physiology.
The potential of magnetic resonance imaging in providing information on lung volume needs also to be considered, as this provides three dimensional image data without incurring any risk due to ionizing radiation. Conventional proton MR does not provide images that are easily related to air volume, and so are unlikely to compete with CT for this purpose. However, measurements of regional lung volume would be clearly possible using hyperpolarised gas imaging [
37]. If this were to become readily available then it could find an important role in this area.
Competing interests
This work was part of study (EudraCT # 2007-003563-43) which was sponsored by Air Liquide. JF acts as a consultant to Air Liquide. The University of Southampton has received funding from Astra Zeneca, Sweden for a study in which JF, JC and MB were involved.
Authors’ contributions
JF designed the study, developed the image analysis software used in the study and drafted the manuscript. JC was the principal investigator of the overall project of which the current study was a part, and oversaw the clinical acquisition of the imaging data. CM performed the image analysis. MB and SM contributed to the design of the software. GC contributed to the study design. IK contributed to the study design and helped to draft the manuscript. All authors read and approved the final manuscript.