Study of the structure of the air and blood capillaries of the gas exchange tissue of the avian lung by serial section three-dimensional reconstruction



    1. School of Anatomical Sciences, Faculty of Health Sciences, University of the Witwatersrand, 7 York Road, Parktown 2193, Johannesburg, South Africa
    Search for more papers by this author
  • J. N. MAINA

    1. School of Anatomical Sciences, Faculty of Health Sciences, University of the Witwatersrand, 7 York Road, Parktown 2193, Johannesburg, South Africa
    Search for more papers by this author

J. N. Maina. Tel: +27-011-717-2305; fax: +27-011-717-2422; e-mail:


We have previously reconstructed the gas exchange tissue of the adult muscovy duck, Cairina moschata using a method of manually aligning sections and tracing the contours of the components of the gas exchange tissue. This reconstruction method demonstrated that the air capillaries are comprised of an expanded globular part interconnected by narrow air channels. The blood capillaries completely surround the air capillaries forming an anastomosing meshwork of short segments. However, the resulting reconstruction was limited in scope because of the laborious process of tracing the profiles of each component through the sequence of micrographs. We have now reconstructed a larger proportion of the exchange tissue by using a cross-correlation based alignment strategy and have demonstrated that the staining intensity of each of the exchange tissue components is sufficiently different to allow them to be identified by simple filtering and thresholding. The resulting reconstructions sample a much larger proportion of the exchange tissue and demonstrate the heterogeneity of structures from different locations in the parabronchus. We have shown that a sheet-flow-type arrangement of blood capillaries surrounds the infundibulum; this represents an unexpected functional convergence with the arrangement of blood capillaries surrounding the mammalian alveoli. It is feasible, using this reconstruction strategy, to analyse the exchange tissue of a large number of avian species in order to determine structural correlates of function. The resulting reconstructions could be analysed in order to determine the basis of the functional efficiency and rigidity of the avian lung.


To understand the function of complex organ systems, the shapes, sizes and the spatial arrangement(s) of the constitutive structural components need to be determined at the highest possible level of resolution. In the case of the mammalian respiratory system, a simplified model of the acinus, the gas exchange unit, has been proposed by Kitaoka et al. (2000) and has subsequently been used to develop a model of oxygen diffusion and absorption in the lung (e.g. Felici et al., 2002, 2004). Although the mammalian lung has been studied extensively by a variety of methods, details of the three-dimensional (3-D) morphology and arrangement (spatial configurations and contacts) of the terminal gas exchange units of the avian lung, the air capillaries (ACs) and the blood capillaries (BCs), still remain unclear (e.g. King & McLelland, 1989; López et al., 1992; Maina, 2005).

Compared to that of mammals, from its exceptional structural complexity (e.g. Maina, 2005, 2007), the avian lung presents a great challenge. Totally rigid, the lung is ventilated continuously and unidirectionally in a caudocranial direction by synchronized bellows-like action of two sets of air sacs, a cranial and a caudal group (e.g. Scheid, 1979). From the parabronchial lumen, the air flows outwards into the atria and infundibulae and from there diffuses into the ACs (e.g. Crank and Gallanger, 1978). Blood flowing from the periphery, i.e. from the interparabronchial artery, the interparabronchial arterioles, and ultimately into the BCs meets the air moving centrifugally (outwards). A cross-current design is formed between the air flowing in the parabronchial lumen and the deoxygenated (venous) blood flowing inwards from the periphery of the parabronchus whereas a countercurrent system is formed between the air in the ACs (flowing outwards) and the blood in the blood capillaries (flowing inwards) (e.g. Scheid & Piiper, 1972). A satisfactory model of gas exchange in the avian lung should therefore require structural data that at least include a part of the parabronchial lumen, including an infundibulum and all of the ACs.

We have previously (Woodward & Maina, 2005) produced a high-resolution reconstruction of part of this area from the Muscovy duck, Cairina moschata by serial sectioning, manual alignment and manual segmentation. This represents the first 3-D reconstruction of the exchange tissue of the avian lung. Although this reconstruction established the 3-D shape and relationship of the exchange tissue components, the ACs and the BCs, it was limited in scope because of the laborious and time-consuming nature of the reconstruction process that limited the volume and number of reconstructions that could be produced. In general, the manual segmentation process is often the rate-limiting step when analyzing a volume (McIntosh et al., 2005).

Here, we describe an automated reconstruction method using cross-correlation to align serial sections and filtering and thresholding to segment the gas exchange components. The resulting reconstructions are comparable or superior in quality to those produced by manual segmentation and, because of the ease with which reconstructions can be produced we have been able to produce reconstructions from a variety of locations within the infundibulum. This has demonstrated the morphological heterogeneity of the ACs and the BCs across a greater part of the exchange tissue than previously demonstrated, information that is vital for the construction of a model infundibulum, apparently the avian homologue of the mammalian acinus.

Materials and methods

Fixation of the lung, tissue sampling and processing

Three specimens of adult muscovy ducks, Cairina moschata were killed by injection with Euthanase® (200−3 pentobarbitone sodium—Centaur Laboratories, Mumbai, India) into the brachial vein (0.01−1). The lungs were immediately fixed by intratracheal instillation of 2.5% glutaraldehyde solution buffered in phosphate (osmolarity 350 mOsm per litre, pH 7.4) at a pressure head of 3 kPa (25 mmHg). The trachea was ligated and the fixative left in situ for 3 h before the lungs were carefully dissected from the deep costal attachments. Subsequently, the lungs were sliced transversely along the costal sulci at approximately 10-mm intervals. The slices were laid out flat and cut into cubes of about 2 mm3. The lumina of the parabronchi were visible as small holes running through the sampled tissue blocks. Those samples that contained parabronchi that were transversely cut were selected. Five pieces were picked from the lungs of each of the three specimens of birds. The pieces were left in 2.5% glutaraldehyde at 4°C overnight and then postfixed in 1% osmium tetroxide buffered in 1 mol per litre sodium cacodylate (osmolarity 350 mOsm per litre, pH 7.4) at room temperature for one-and-half hours. This was followed by dehydration in a series of concentrations of ethanol starting from 70% to absolute followed by two changes of propylene oxide before embedding in epoxy resin (epon/araldite).

One hundred ninety four 0.3-μm thick serial sections from the exchange tissue were cut and stained with toluidene blue. These were roughly aligned and photographed at 400× magnification (Axioscope image analyzer, Zeiss instruments, Jena, Germany) as previously described (Woodward & Maina, 2005).

Reconstruction and volume segmentation strategy

A series of 143 micrographs, spanning a depth of ∼43 μm were used in the present study. Images were down-sampled to 0.3 μm per pixel (equal to the z-dimension sampling) and manually corrected (brightness and contrast) using The Gimp V.2.2.13 (GNU general public license) before being exported to raw format using IrfanView 3.98 (Copyright Irfan Skiljan). The images were imported into Spider V.13 (Frank et al., 1996) and normalized to a mean of 0 and a standard deviation of 1. Several areas of interest were selected using Web (Frank et al., 1996); these were aligned with subsequent sections by normalized cross-correlation (e.g. Rath & Frank, 2003; Roseman, 2003). In real space, the search window would be ‘superimposed’ on the micrograph at every point in turn. At every position a normalized correlation coefficient is obtained between the pixel densities in the two images. The search window is rotated by 5 degrees and the process is repeated. The coefficient with the highest value represents the ‘best fit’ between the search window and the next micrograph (Roseman, 2003). In practice, the coefficients are calculated in Fourier space to decrease search time. The process of alignment is illustrated in Fig. 1.

Figure 1.

The algorithm used for aligning serial sections. An area of interest (n) is selected from the first (686 × 686 pixel) micrograph; this is padded and cross-correlated with the next micrograph in the image sequence (n+ 1). The peak correlation value, which represents the x and y translations and in-plane rotation necessary to maximally align the area of interest in the next micrograph, is identified. These values are then used to window the micrograph, thus producing a new search window (n+ 1), which is fed back into the algorithm. This process is repeated for every image in the micrograph stack resulting in an aligned image sequence. The area of interest is selected using Web (Frank et al., 1996); windowing, rotation and cross-correlation are achieved using Spider V.13 (Frank et al., 1996).

The aligned image sequences were combined into volumes of variable dimension and low-pass Fourier filtered to 3-μm resolution using a Gaussian fall-off. This eliminated high-frequency noise such as specks of dirt, differences in section intensity and misalignment of sections. A copy of the volume was contrast-inverted and both the original and the inverted copy were opened simultaneously in Chimera (Pettersen et al., 2004). Two threshold levels were selected (Fig. 2) with every voxel falling below the lower threshold being assigned to the class of blood capillaries and all of those pixels above the higher threshold falling into the air-capillaries class. Voxels lying between the two threshold levels were not assigned. The optimal thresholding and filtering parameters were estimated by comparing the profiles of the structures of interest obtained by this procedure with the original sections (Fig. 2). Because the images had been normalized to the same mean and standard deviation, these parameters were successfully applied to reconstructions taken from different locations on the micrographs.

Figure 2.

A single unfiltered and aligned section is shown on the lower left; air capillaries (AC) and blood-capillaries (BC) have been manually identified. To the right is the same section after Fourier filtering. The labels from the left-hand section (AC and BC) are superimposed onto this section as well as on two threshold levels. The upper threshold lying at a grey-scale intensity level of ∼157 defines the lower range of the set of air capillary voxels. The lower threshold lies at a grey-scale level of ∼148 and defines the upper intensity range of blood-capillary voxels. A horizontal line has been drawn across the centre of the section and the pixel intensity profile is plotted above. All voxels occurring above the upper intensity level are air capillaries, and below are the lower, blood capillaries. This allows automatic identification of the profiles of the components of the gas exchange tissue. This resulting contours correlate well with the manually identified structures.

The various reconstructions were rotated, sliced and magnified in an interactive way. The gas exchange components; the air and blood capillaries were separated from one another and then merged to determine the exact architectures of the components and interactions throughout the entire infundibulum. Rendering and volume manipulation were done using the UCSF Chimera package from the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco (supported by NIH P41 RR-01081) (Pettersen et al., 2004). The reconstruction of an averaged sized region (∼200 voxels3) takes approximately 1 h, half of which is spent in the automated cross-correlation routine. Computation was performed on one 2.2 GHz Xeon processor on an Intel server with 1 GB RAM under the Fedora Linux operating system. For comparison purposes, the manual method requires approximately 140 man-hours for a comparable sized region.


The reconstruction preserves sufficient resolution for the individual exchange tissue components, the air and blood capillaries, to be clearly identified (Fig. 3). The blood-gas barrier is not resolved; this is most clearly seen in places where two air capillaries or two blood capillaries are separated from each other (Fig. 3). This was likewise the case for the reconstruction resulting from manual alignment and segmentation (Woodward & Maina, 2005). The automated alignment and segmentation method was applied to the sections previously reconstructed manually. A comparison between the two reconstruction methods demonstrates that the automated method described here correctly aligns the sections and segments the air and blood capillaries as well or better than the manual method (Fig. 4). Furthermore, the manual method was limited in scope because of the laborious segmentation method employed, to a depth of ∼21 μm (Fig. 4). The automated method has allowed us to reconstruct a sample of the parabronchus spanning the entire exchange tissue from the parabronchial lumen to the parabronchial septum for the first time (Fig. 5).

Figure 3.

(A) The first section in an image stack, a few air capillaries have been manually identified and labelled with letters. (B) A reconstruction of the area shown in (A) and labelled in an identical way, the air capillaries are shown in blue and the blood capillaries in red. The profiles correlate well, except that the blood-gas barrier -is not resolved. (C, D). The blood (C) and the air capillaries (D) can be separated from one another in order to better understand their structure and arrangement. Scale bars indicate 3 μm.

Figure 4.

A comparison between a manually segmented and aligned reconstruction and a reconstruction of the equivalent area resulting from automated alignment and segmentation. (A) The manually reconstructed air capillaries (AC) and the automatically reconstructed air capillaries (B): the general morphology of the two reconstructions correlate well. (C) The manually reconstructed blood capillaries and the automatically reconstructed air capillaries (D). In the automatically reconstructed blood capillaries (D), (*) indicate areas incorrectly assigned to the set of blood conducting elements in the manual reconstruction (C). The automated reconstruction has an open structure, more similar to that produced by corrosion casting of these structures (Maina 1982, 1988). The entire extent of the manual reconstruction is 21 μm.

Figure 5.

A reconstruction spanning the parabronchial exchange tissue. (A, B) Comparison of the reconstruction (A) with the original exchange tissue (B) The parabronchial lumen (PL) can be seen giving rise to the atrium (At), which in turn gives rise to the infundibulum (If). A few air and blood capillaries have been labelled (AC, BC). The parabronchial septum is just outside of the reconstructed area in the top right-hand corner. (C, D) The air spaces and blood vessels have been isolated from one another. Narrow air channels (*) can be seen interconnecting the globular air capillaries, these are labelled in both reconstructions. The blood capillaries (BC) form an interconnected open-meshwork structure. AC, air capillaries; AV, atrial vein. Scale bars indicate 50 μm.

The infundibulum

The infundibulum projects into the exchange tissue from the parabronchial lumen (Figs. 5 and 6A). It is surrounded by a network of interconnected blood capillaries that become visible if we remove the air-conducting elements (Fig. 6B). Narrow air channels project through the meshwork of blood capillaries and pass through the openings shown (Fig. 6B). Removing the blood capillaries reveals that narrow air channels project from the infundibulum and pass through the blood capillary layer (Fig. 6C). At higher magnification the blood capillary meshwork is seen to be almost flat, forming a sheet-flow-type arrangement. The sheet is perforated by multiple holes that allow the narrow channels of the air capillaries to pass through. The individual blood capillaries are fairly shorter than their width (Fig. 6D, E, F).

Figure 6.

(A) The infundibulum (If) surrounded by blood capillaries can be seen in this reconstruction showing both the air- and blood-conducting elements. (*) indicates a narrow air channel. (B) The air-conducting elements have been removed leaving behind the blood capillaries, which frequently anastomose forming a dense, flat meshwork. The space left behind by the infundibulum is labelled (If). Holes in the side of this space, the exit points of the narrow air channels, are indicated by arrows. The circle indicates an area that is magnified and tilted in (D). (C) The air-conducting elements with blood capillaries removed; the narrow air channels visible as empty spaces in (B) are obscured here by the infundibulum. The surface of the infundibulum is flat and gives rise to narrow air channels (*). (D) The area indicated by arrows in (B) and (D, E, F) are the holes left by the exiting air channels. The blood capillaries whose diameter fairly exceeds their length are indicated (BC). The blood capillaries form a surface matrix, completely surrounding the infundibulum. These blood capillaries (D) are narrowed in places (#) where they lack erythrocytes and can be seen to be narrowed in toluidine blue stained sections. Scale bars indicate 3 μm.

The narrow air channels connect the globular air capillaries to the infundibulum (Fig. 7A). Removing the blood capillaries allows us to visualize the surface of the infundibulum giving rise to air channels (Fig. 7B). Again, from this view, the constricted holes in the blood capillary network through which the air channels pass, are visible (Fig. 7C). At higher magnification, the narrow air channels can be seen giving rise to the globular air capillaries. These, in turn, are interconnected with other air capillaries by narrow channels (Fig. 7D–F).

Figure 7.

(A) The infundibulum (If) is surrounded by a blood capillary layer. (B) The blood capillaries have been removed and a few of the narrow channels projecting from the infundibulum are visible (*). The surrounding space is filled with blood capillaries which pass through the spaces that are indicated by arrows. The area enclosed in the square is enlarged in area (D). (C) The blood capillaries with air capillaries removed, arrows indicate where the narrow air channels perforate the surface formed by the blood capillaries. (D, E, F). Magnified views of the infundibulum giving rise to the narrow air channels (*), which expand to form the globular air capillaries (AC). Multiple narrow air channels (arrows) project from these irregular structures and interconnect with one another. Scale bars indicate 3 μm.

The exchange tissue

The air and blood capillaries that are not in a direct contact with the infundibulum are distinct, morphologically, from those surrounding or projecting directly from the infundibulum. The air capillaries are rotund structures, displaying a variety of forms and number of interconnecting narrow air channels (Fig. 8C, D). They completely surround the blood capillaries, which form a three-dimensional profusely anastomosing meshwork (Fig. 8A, B).

Figure 8.

(A, B) Magnified views of the blood capillaries that are not in a direct contact with the infundibulum. Arrows indicate the spaces where the narrow air channels pass through. (*) indicates those blood capillaries that lack erythrocytes. The blood capillaries form a three-dimensional meshwork of interconnected capillary segments. (C, D) Air capillaries with no direct continuity with the infundibulum, the air capillary expansions are connected by a variable number of narrow air channels (*). A blood capillary passes through the space indicated with an arrow. Scale bars indicate 3 μm.

Dimensions of the air and blood capillaries

The globular part of the air capillaries measures between ∼3 and 30 μm in diameter. The interconnecting air channels range in length from ∼2 to ∼4 μm and in diameter from ∼0.5 to ∼2 μm. The blood capillaries are uniform in diameter across the parabronchus; their lengths approximate their diameters ∼3 μm, except in cases where the air capillaries are narrowed and devoid of erythrocytes, in these cases the diameter is ∼1 μm.


Comments and critique of the methods

Alignment: During the process of cutting, mounting and photographing serial sections; the orientation of a given section relative to the preceding section in the original sample is lost. The section has three degrees of freedom; in-plane rotation about the vertical axis and translation along the x and y directions. Previously, we used a manual realignment strategy that involved rapidly switching between sections and applying rotations and translations to minimize apparent movement. This time a 3-D, normalized cross-correlation function was calculated between a template; consisting of a small area of the preceding section and the section to be aligned (Fig. 9). The maximum correlation values represent the alignment operators that, when applied to the section, minimize the differences between it and the template (reviewed by Fiala, 2005). The resulting alignment was inspected by viewing the volume as a movie.

Figure 9.

A typical cross-correlation function, calculated between an area of interest and the next micrograph. On the z-axis (also shown by hue) is the correlation coefficient. Figures A–D show the effect of rotation, the correct rotation, required to bring the area of interest into register with the next serial section is −26°, shown in (C), this rotation results in the highest relative correlation coefficient (for the normalized correlation coefficients associated with each peak see Fig. 10). The x and y translations are taken from the position of the peak (arrows) in (C). A peak of comparable size (*) to the correct one can be seen when the rotational alignment is incorrect (A).

Because the cross-correlation is calculated locally, over the limited dimensions of the reconstructed area, distortions, resulting from shearing or compression of the entire section (e.g. Machin et al., 1996), are minimized (Stevens et al., 1980). We also found this to be the case during our previous reconstruction (Woodward & Maina, 2005). We attempted the use of a correlation threshold to identify cases where sections were folded or torn in the area to be reconstructed; unfortunately this approach was unsuccessful (Fig. 10). Incorrect matches had to be manually detected by watching the volume sequence as a movie. Any section that results in an anomalously low maximum correlation coefficient is eliminated from the reconstruction and the preceding section is duplicated. This approach was also taken with missing sections.

Figure 10.

The normalized correlation coefficient calculated between the area of interest and the next serial section in the series. The correlation coefficients are generally low (<0.3) however, they are clearly distinguishable from the background noise (see Fig. 9). Between sections 65 and 66 an incorrect match was made, however, because the associated correlation coefficient was within normal limits, a threshold value could not be used to spot this mistake. This means that once the alignment algorithm is complete, the image sequence should be viewed in a rapid succession so that incorrect matches such as these can be manually corrected.

Segmentation: In order to interpret the volume, aligned series of sections can be viewed rapidly in a sequence. This gives the impression of moving through the tissue, by moving backwards and forwards repeatedly an impression of the structures of the air and blood capillaries is gained. However, a more powerful approach is to segment the reconstruction by identifying structures of interest, which can be fitted with a smooth mesh and coloured. Our approach used the intensity difference between the lumen of the air capillaries compared to the other structures (blood-gas barrier, plasma and erythrocytes). The intensity differences between sections were minimized by adjusting each micrograph using The Gimp V.2.2.13 and normalizing the images in an automated fashion using Spider v.13 (Frank et al., 1996).

Any voxel, lying above the upper threshold value was assigned to the set of air capillaries. The blood capillaries are more difficult to identify as, there is no way to distinguish between the intensity of the plasma and blood-gas barrier, which are similar. Therefore, after Fourier filtering, which averages adjacent voxels, those voxels of lower intensity (plasma or blood gas barrier) lying in close proximity to the air capillaries take on intermediate intensity values and are not assigned to either the class of blood or air capillaries. Those lying above this lower threshold value, lie further from the air capillaries and therefore form part of the set of blood capillary voxels.

The shortcoming of this method is that some detail is lost, most importantly in cases where two air or two blood capillaries lie adjacent to one another, separated by a tissue barrier, which is not resolved. This has the effect of making these appear connected, when they are in fact isolated from one another. However, as in the case of our previous reconstruction (Woodward & Maina, 2005) we believe that this has minimal effect on the reconstructions because of the open arrangement of the air and blood capillaries. The reconstructions produced using this method resemble those produced by corrosion casting (Maina, 1982, 1988) and those produced by manual segmentation and confocal laser scanning microscopy (CLSM) by Woodward & Maina (2005). Cases in which air capillaries contain erythrocytes were rare, but where these are present they can easily be identified and interpreted accordingly, particularly if the adjacent blood capillaries contain blood plasma.

Comparison with previous findings

The 3-D morphology and organization of the components of the terminal gas exchange tissue of the avian lung has been studied previously by corrosion casting (Maina, 1982, 1988), CLSM and serial section reconstruction (Woodward & Maina, 2005). Corrosion casting followed by scanning electron microscopy provides insight into 3-D structures, but comes with important shortcomings. Foremost among these is that structures such as blood vessels need to be perfused first and therefore lose their normal shape and size; moreover, artefacts may arise from the use of excessive or insufficient pressure when applying the casting material. However, when correctly interpreted, the resulting images resemble those produced by serial section reconstruction (Woodward & Maina, 2005). The limitation of manually aligning and segmenting serial section image stacks is the limited scope and number of reconstructions possible because of the time-consuming nature of the reconstruction process. The automated method described here, by contrast, has facilitated the production and analysis of a much greater proportion of the gas exchange tissue. It has been possible to reconstruct volumes sampled at positions from every point of the parabronchial exchange tissue; from the parabronchial lumen to the interparabronchial septum. The resulting structural insights are made with the confidence that comes from having a larger sample size.

The two reconstructions were compared by superimposing them in three dimensions; this is illustrated in Fig. 4; where the two reconstructions are shown in the same orientation. The differences between the two are slight; but, especially in the case of the blood capillaries, the arrangement of the respiratory units demonstrated by the two reconstruction methods is different in some areas (Fig. 4C, D). It is important to ask the question ‘which reconstruction is a closer approximation of the actual tissue structure?’ and if we take the painstakingly contoured and aligned manual reconstruction as a gold standard and compare the new method to this we conclude that the automated reconstruction provides the same overall picture of the terminal gas exchange tissue, but at a slightly lower quality. However, if we compare the goodness of fit of both reconstructions in a subjective way with the morphology demonstrated by corrosion casting followed by SEM (Maina, 1982, 1988) and our CLSM reconstruction of the blood capillaries (Woodward & Maina, 2005), the open arrangement demonstrated by the automated procedure appears to fit more closely. The reasons for this are not clear.

Physiological models, devised to demonstrate the physical basis of lung function have been produced. However, these have been based on inexact or incorrect terminal gas exchange tissue morphology: Scheid (1978), Crank & Gallagher (1978), Powell (1982) and Shams & Scheid (1989) assumed that the air and blood capillaries were straight, non-branching tubules. Brackenbury & Akester (1978) predicted that the air and blood capillaries were mirror images of one another. The actual arrangement and forms of the terminal respiratory units differ substantially from these assumptions. The blood capillaries that are not in a direct contact with the infundibulum form a volume-flow arrangement, completely surrounding and intertwining with the globular air capillaries that are interconnected by narrow air channels. This volume-flow arrangement has been observed previously by Maina (1982, 1988) and Woodward & Maina (2005). Interestingly, the blood capillaries form a sheet-flow arrangement (Maina, 2000), surrounding the infundibulum. This has not previously been reported in the gas exchange tissue of the avian lung and shows morphological convergence with the interalveolar septal capillary system of the mammalian lung.


It has previously been demonstrated that the terminal respiratory units of the avian lung; the air and the blood capillaries anastomose profusely and intertwine to form a complex meshwork. The air capillaries consist of globular parts that interconnect via narrow air channels. The blood capillaries consist of short segments that have approximately equal diameters and lengths. By applying automated methods to the problems of realignment of serial sections and segmentation of the volume, a much larger number of reconstructions were possible. We have demonstrated that the infundibulum is surrounded by a sheet-flow-type blood capillary arrangement, analogous to the arrangement surrounding the mammalian alveoli, and produced a 3-D reconstruction of a cross-section of a parabronchus, extending from the parabronchial lumen to the parabronchial septum. This reconstruction could be used as a basis for simulation models of gas exchange and contributes to determining the basis for the efficiency of the avian pulmonary system or be subjected to finite-element analysis to determine the basis for the rigidity of the avian lung.


We wish to thank Mr. R. Tseki, Mrs. S. Rodgers, Mrs A. Mortimer, Dr. H.P. Brunhubner, Mrs. C. Lalkhan and Ms. P. Sharpe for their excellent technical assistance. This work was funded by grants from the NRF (National Research Foundation) and the University of the Witwatersrand, Faculty of Health Sciences Research Committee.