Study of the Variability in Upper and Lower Airway Morphology in Sprague УDawley Rats Using Modern Micro-CT Scan-Based Segmentation Techniques.код для вставкиСкачать
THE ANATOMICAL RECORD 292:720–727 (2009) Study of the Variability in Upper and Lower Airway Morphology in Sprague– Dawley Rats Using Modern Micro-CT Scan-Based Segmentation Techniques JAN W. DE BACKER,1,3* WIM G. VOS,3 PATRICIA BURNELL,4 STIJN L. VERHULST,3 PHIL SALMON,5 NORA DE CLERCK,6 AND WILFRIED DE BACKER3 1 FluidDA, Drie Eikenstraat 661, 2650 Edegem-Antwerp, Belgium 2 Vision Lab, University of Antwerp, Universiteitsplein 1, 2610 Wilrijk–Antwerp, Belgium 3 Department of respiratory medicine, University Hospital Antwerp, Wilrijkstraat 10, 2650 Edegem, Antwerp, Belgium 4 GlaxoSmithKline Research and Development, Park Road, Ware, Hertfordshire, SG12 0DP, UK 5 Skyscan, Kartuizersweg 3B, 2550 Kontich, Belgium 6 Department of Biomedical Sciences, University of Antwerp, Universiteitsplein 1, 2610 Wilrijk–Antwerp, Belgium ABSTRACT Animal models are being used extensively in pre-clinical and safety assessment studies to assess the effectiveness and safety of new chemical entities and delivery systems. Although never entirely replacing the need for animal testing, the use of computer simulations could eventually reduce the amount of animals needed for research purposes and reﬁne the data acquired from the animal studies. Computational ﬂuid dynamics is a powerful tool that makes it possible to simulate ﬂow and particle behavior in animal or patient-speciﬁc respiratory models, for purposes of inhaled delivery. This tool requires an accurate representation of the respiratory system, respiration and dose delivery attributes. The aim of this study is to develop a representative airway model of the Sprague–Dawley rat using static and dynamic micro-CT scans. The entire respiratory tract was modeled, from the snout and nares down to the central airways at the point where no distinction could be made between intraluminal air and the surrounding tissue. For the selection of the representative model, variables such as upper airway movement, segmentation length, airway volume and size are taken into account. Dynamic scans of the nostril region were used to illustrate the characteristic morphology of this region in anaesthetized animals. It could be concluded from this study that it was possible to construct a highly detailed representative model of a Sprague–Dawley rat based on imaging modalities such as micro-CT scans. C 2009 Wiley-Liss, Inc. Anat Rec, 292:720–727, 2009. V Key words: small animal imaging; micro-CT; rat; respiratory tract; 3D model Grant sponsor: GlaxoSmithKline. *Correspondence to: Jan W. De Backer, University Hospital Antwerp, Department of respiratory medicine, Wilrijkstaat 10, 2650 Edegem, Antwerp, Belgium. Fax: þ32 3 821 44 47. E-mail: Jan.DeBacker@ua.ac.be C 2009 WILEY-LISS, INC. V Received 4 July 2008; Accepted 11 December 2008 DOI 10.1002/ar.20877 Published online 25 March 2009 in Wiley InterScience (www. interscience.wiley.com). AIRWAY MORPHOLOGY IN SPRAGUE–DAWLEY RATS Animal models are being used extensively in pre-clinical and safety assessment studies to assess the effectiveness and safety of new chemical entities, delivery systems or to investigate the safe levels for occupational exposure (Alfaro et al., 2004; Chapman et al., 2007; Ji et al., 2007). Often small animals, like mice or rats, are exposed to the product for a longer periods of time (Wagner et al., 2006; Lee et al., 2007) or for shorter periods of time to assess acute effects (Dong et al., 2005; Kouadio et al., 2005). As in human trials (De Backer et al., 2007a) a trend is emerging in animal studies towards accurate computer models describing the animals’ physiological behavior (Schroeter et al., 2006; Galle et al., 2007). Although never entirely replacing animal studies, this could, eventually, reduce the amount of animals needed for research. The increasing prevalence of respiratory diseases such as asthma and chronic obstructive pulmonary disease (COPD) intensiﬁed the research regarding the morphology and physiology of the respiratory system both clinically as well as pre-clinically. A tool that is increasingly used to assess the respiratory function is computational ﬂuid dynamics or CFD. This method is capable of simulating ﬂow behavior in virtual models (Bush et al., 1998; Andersen et al., 2000; Minard et al., 2006). The challenge is to develop mathematical models that realistically reﬂect the respiratory physiology of the animal by accurately simulating its ﬂuid ﬂow conditions. These models comprise of accurate geometrical representations of the airway structures and realistic boundary conditions which drive the ﬂow (De Backer et al., 2007b). The CFD method and these models can assist in the development of new inhalation therapies and the optimization of inhaler devices (Coates et al., 2007; Kleinstreuer et al., 2007). Advances in small animal imaging using micro-CT or MRI have made it possible to study the respiratory system in a speciﬁc animal into great detail (Johnson, 2007; Lam et al., 2007; Wietholt et al., 2008). Earlier work by Chaturvedi and Lee (Chaturvedi and Lee, 2005) has used micro CT images to accurately study the morphology of the mouse and canine airway tract based on lung casts and segmentation principles. The study we present here uses similar techniques but scans are performed in vivo or in situ. Our study aims to develop a representative, realistic model of the rat respiratory tract based on high-resolution micro-CT scans. In particular the variability between different animals from the same species will be investigated as also performed in another recent study by Lee et al. (Lee et al., 2008). Lee et al. again used lung casts to assess the geometrical features of the respiratory system of healthy Sprague–Dawley rats. In our study, a range of techniques (static and dynamic micro-CT scans) will be used in different circumstances (anaesthetized and euthanized animals) to obtain in vivo and in situ images. Differences between different rats were assessed to obtain a representative model. Both the nasal passages and the tracheobronchial tree were considered in the model. The resulting representative model could subsequently be used in ﬂow and particle deposition simulation studies. MATERIALS AND METHODS In this study a total of 11 Sprague–Dawley rats were used. The average weight of the animals was 372 56 g. Micro-CT scans (Skyscan 1076 high resolution in vivo 721 micro-CT, Skyscan, Belgium) were taken of the snout and entire respiratory tract of the rats. The scanner had a tube diameter of 68 mm and a 17 mm single scan length. For the scans a resolution of 35 lm was selected. An initial test comparing scans from anaesthetized with euthanized rats clearly showed that resolution of airway structures in the euthanized rat was superior to the anaesthetized rat. Therefore, the airways in the lungs were scanned in 7 euthanized rats to optimize the image quality. The tracheobronchial tree was deﬁned as the airways starting from the larynx down. The nasal passages were scanned in a total of 8 rats. A deﬁnition of the airway regions is given in Fig. 1 and an overview of the scans is provided in Table 1. In addition to the volume of the lower airways, the airway length was determined through the construction of centerlines based on the segmentation masks lCT Scanning of the Respiratory System The ﬁrst rats (rat 1–4) were scanned from the nostrils down to the diaphragm to include the lungs and all extrathoracic structures (snout, nares, nasopharynx, epiglottis, and trachea) The total scanning length was 12 cm (Fig. 1) and required around 6–7 scans per animal. Scans were taken with a source voltage of 100 kV and a source current of 140 lA. The resolution was set to 35 lm and the rotation step was 0.6 degrees. The rats needed to be euthanized for these scans because the scanning time would have taken too long (>5 hr) to bring animals under anesthesia and to preserve image quality due to motion artifacts. The animals were euthanized with 1 mL of Nembutal and were left for 1 hr to complete the state of rigor mortis. To increase the number of animal-speciﬁc tracheobronchial morphologies, only the tracheobronchial regions of 3 additional rats (rats 5–7) were scanned additionally. Dynamic lCT Scanning of the Nasal Passages Rats are obligate nasal breathers, using their sense of smell to identify and classify objects. Consequently, the nasal passages are well developed and have an important ﬁltering function to prevent hazardous particles from entering the animals’ respiratory system. To assess the movement of the nostril and upper airway region during breathing, dynamic scans of this region were taken in 4 rats (rats 8–11) in addition to the static lCT scans of rats 1–4. Scans were taken with a source voltage of 70 kV and a source current of 140 lA. The resolution was set to 35 lm and the rotation step was 0.9 degrees. The scan length was 1.5 cm. The rats were anaesthetized and their breathing pattern was monitored. The dynamic scans were performed using respiratory gating. The animals were anaesthetized with a 5050 mixture of Nembutal and saline. The dosage was based on 35 mg of Nembutal per 2.5 kg of bodyweight. The respiratory gating was based on a visual triggering system. The motion of the animal’s thorax was gated and whenever a clear signal was detected four images were taken and the registration time was recorded (Fig. 2). Afterwards all images were sorted into several bins to attain static images at certain breathing levels. Therefore images of the entire region of interest where obtained for the normal inspiration and expiration. 722 BACKER ET AL. Fig. 1. Deﬁnition of scanning length for rat airway model. Combining these static images produced a dynamic image, indicating the changes over time. Airway Segmentation All scans from both anaesthetized and euthanized animals were then read into a commercially available and validated software package Mimics (Materialise, Leuven, Belgium). Three-dimensional reconstructions of the upper and lower airways were made, based on segmentation principles. Voxels with Hounsﬁeld Units, a measure of electron density, within the pre-described range (1,024 to 865) were placed in a separate mask. A three-dimensional virtual representation of the structure was reconstructed from this mask. Airway lengths, surfaces, and volumes could be measured from this reconstruction. The selection process for the ‘average animal model’ is explained in the results and discussion section. Segmentation into the various lung lobes was carried out using the ‘average model’. This was done by segmenting the lung volumes and subsequently deﬁning the ﬁssures separating the lung lobes. Ethical Approval This study has been approved by the ethical committee of the University of Antwerp. Animals were handled by the qualiﬁed staff of the animalarium of the University of Antwerp. RESULTS Lower Airway A visual representation of the reconstructed lower airways of rat 1–7 can be found in Fig. 3. The typical structure of the airway tree can be clearly seen in Fig. 3. Measurements of the segmented lower airway volume gave an arithmetic mean airway volume and standard deviation of 332.40 63.9 mm3. The lengths of all centerlines of all branches were added to attain the total airway length. An illustration of the centerlines is given in Fig. 3. The average lower airway length was 174.49 37.96 mm. By dividing the airway volume by the airway length one attained an average cross sectional area (CSA) of the airways. On average this value was 1.95 0.40 mm2. All values for rats 1–7, in which the tracheobronchial tree was scanned, can be found in Table 1. As indicated in Fig. 3 the detected tracheobronchial tree in each rat is slightly different in terms of volume and the number of airways. To investigate the variability between the models a correlation was made between the average CSA and the airway length of a model (Fig. 4). Rats 1–7 are shown in Fig. 4 and the CSA values towards the origin are those of the mean trachea values for all assessed animals. 0.19 0.21 0.24 0.26 – – – 0.12 0.11 0.11 0.11 100 294 119 138 – – – 87 63 68 86 2.6 2.1 1.7 1.4 1.9 1.7 2.1 – – – – 530 1420 486 538 – – – 752 560 601 759 (mm) (mm3) (mm2) (mm2) Nasal passages volume Average cross sectional area 723 An exponential correlation was found with equation with a coefﬁcient of variation r2 ¼ 0.9789. Average CSA ¼ 4:4258 e0:0042ðAirway LengthÞ (1) This correlation was made without the outliers as explained in the discussion section. If the correlation was valid, this would imply that if one would take a segmented model and subsequently cut this model to reduce the segmented airway length, the resulting average CSA would be predicted by the correlation equation. To test this, the model of rat 3 which was the one with the largest volume was cut twice. It was found that each time the resulting average CSA was correctly given by the correlation equation. This is indicated by the crosses in Fig. 4. From this analysis and this correlation it could be concluded that the model that was segmented the furthest was most suitable to represent the rats within this weight category. In addition, small changes in weight within the size range 270–450 g have little impact on lung size. More detail can be found in the Discussion section. 126 155 232 184 208 138 179 – – – – 332 331 386 265 398 231 384 – – – – – – – – – – – X X X X (mm) (mm3) Dynamic lCT For the extra-thoracic airway geometries a slightly different approach is taken. In Fig. 5 the morphology of the nares is presented for the euthanized rats using a dynamic scanning method. The average volume of air in the nares of the euthanized rats is 162.75 88.86 mm3 and for the anaesthetized rats, 75.93 12.27 mm3. The centerlines are less informative for this part of respiratory tract, because these centerlines would be calculated in the complex olfactory and sinus regions that only account for a small fraction of the air mass ﬂow rate. Therefore surface area was calculated to better represent the nares. The euthanized rats had an average surface area of 743.5 451.58 mm2 and the anaesthetized rats, 668 102.45 mm2. Division of the volume by the surface area gives a characteristic length of the nares. For the euthanized rats this characteristic length is 0.22 0.03 mm on average, whereas for the anaesthetized scanned animals the average of this length is lower at 0.11 0.001 mm. The detected characteristic length appears smaller in the anaesthetized animals, with a lower variability. The ‘Average’ Model X X X X – – – – – – – lCT Tracheobronchial tree volume A representative average model was selected using the criteria described in the discussion section. The nasopharyngeal volume of the representative model was 373 mm3, the tracheal volume was 173 mm3 and the volume of the central airways amounted to 233 mm3. The resulting model is shown in Fig. 6. Weight (g) 450 450 320 270 390 380 430 350 330 340 380 Rat 1 2 3 4 5 6 7 8 9 10 11 X X X X X X X – – – – Measurement of Lung Lobes lCT Nasal passages euthanized Nasal passages anaesthetised Tracheobronchial tree length Upper Airway Tracheobronchial tree euthanized TABLE 1. Overview of scanned rats, volumes, lengths, and cross sectional areas Nasal passages total area Nasal passages char. length AIRWAY MORPHOLOGY IN SPRAGUE–DAWLEY RATS A rat has four lung lobes in the right lung and one lobe in the left lung (Greene, 1970) which can be distinguished in the micro-CT images by following the ﬁssure lines. One lobe of the right lung is physically present in the left part of the rat’s thorax. Air is supplied to this lobe through the characteristically large airway which laterally crosses the thorax. Lung lobe measurements are shown in Fig. 6, where the airways are numbered and colored according to the lobe they provide with air. Figure 7 presents the values for the volumes and surface areas of the lung lobes with the nomenclature as deﬁned by Lee et al. (2008). 724 BACKER ET AL. Fig. 2. Respiratory gating in dynamic scans of rat upper airway. Fig. 3. Rat tracheobronchial trees (ventral view) en illustration of centerlines. DISCUSSION Drug delivery through inhalation has become a popular method the past decades to attain local therapeutic effects. The inhalation route could also be an attractive alternative to attain systemic effects because for the same amount of medication direct delivery to the lungs results in a higher biological availability compared to oral medication. However, because adequate technologies AIRWAY MORPHOLOGY IN SPRAGUE–DAWLEY RATS 725 Fig. 4. Average cross sectional area as a function of airway length. have not been commonly used in the preclinical setting, few investigators have addressed the detailed deposition of drugs in the lung (Ewing et al., 2008). For this reason, developers of new inhalation therapies would beneﬁt greatly from having a representative, virtual rat model of the airways. It is, however, impractical to develop a speciﬁc model for each animal under investigation. Therefore the aim of this work was to select or construct the average model of the rat, obtained through imaging techniques, whereby the attributes of candidate formulations or delivery systems may be applied to assess deposition within its respiratory tract The primary aim of the project was to analyze several animals in the given weight category and subsequently select the average sized structures, giving an average animal. The average animal should contain as much information as possible on the airway anatomy and at the same time must be representative for the species in this category. There are signiﬁcant issues in determining the attributes of the ‘average’ rat; namely whether the model should be based on the average volume, the average length or the average surface area. Other considerations include the effect of image quality on the segmentation procedure. The extent to which a model could be reconstructed from the segmented images depends mainly on the amount of noise in the images. The noise level can be inﬂuenced by the size of the animal because fat pads attenuate the signal. Therefore selection on the basis average volume or area of the models could result in a nonoptimal model. From Fig. 4, it can be seen that two models (4 and 6, indicated by the solid triangular symbols) appear to be outliers, that is, for the measured airway length the average CSA appears to be smaller than for the other models. Inspecting the 3D reconstructed models and the micro-CT data revealed that these models appeared to be deﬂated during the scan which lasted 6 hr. The rats were not intubated during the scans as this would disturb the geometry of the trachea and the upper airway. Therefore it appears that there is a correlation between segmented airway length and the average CSA for all rats, except 4 and 6, where there was signiﬁcant deﬂation in the airways. The correlation shown in Fig. 4, excluding the outliers, shows that 98% of the variability in the data can be explained by the variation in the segmentation length. On the assumption that the errors involved in the segmentation process were consistent and negligible, the variability between rats, in this weight range and detected volume of the tracheobronchial tree, is around 2%. As a result, it is logical to select the model that was segmented the greatest length. Consequently, this was also the model with the largest tracheobronchial volume and not the average lower airway volume. Important to note is that, unlike in human airways, the rodent airways do not have a dichotomic branching structure, 726 BACKER ET AL. Fig. 5. Rat nasal passages. but rather a central bronchus. This can have a signiﬁcant effect on lung function and particle deposition as will be investigated in a future study. Having selected a representative tracheobronchial airway model, the most suitable upper airway model must now be selected. As can be seen in Fig. 5 some differences exist between the upper airway models attained in the euthanized animals and the models from the anaesthetized animals. In the so-called ‘nasal cavity’ region, indicated in Fig. 5, the differences are more apparent. In the dynamic scans taken while the rat was breathing, this very narrow region shows a small oscillatory dilating movement. It appears to be that the airways in this region collapse once the animals are euthanized. This is conﬁrmed by the calculation of the characteristic length. For the anaesthetized animals this value is almost constant with a very small variability. For the euthanized animals this value is higher and also the variability is larger. To construct a representative model the upper airway morphology must be adapted to reﬂect the behavior as observed in the living animals. Therefore the rat with the best deﬁned nasal geometry was selected to be representative of a ‘living’ rat. By means of Boolean operations the results of the snout and nasal passages from this rat was merged with the scan of the tracheobronchial tree to form an average and representative model of the rat’s respiratory system. CONCLUSION Combining high resolution micro-CT scans with dynamic micro-CT scans made it possible to create a representative virtual geometry of the respiratory system of the Sprague–Dawley rats. It was possible to identify anatomical landmarks and a correlation was found between the segmented airway length and the average of the airways’ CSA. The micro-CT images contained sufﬁcient detail to identify the ﬁssure lines in the lungs, enabling reconstruction and measurement of lobular volumes. The virtual model will now be used to simulate ﬂuid ﬂow conditions during respiration and hence to assess deposition patterns of inhaled particles. AIRWAY MORPHOLOGY IN SPRAGUE–DAWLEY RATS Fig. 6. Mean rat model (left, ventral view), lobular segmentation (top right), and airways colored by lobular pathway (bottom right). Fig. 7. Volume and surface area measurements of the lung lobes for the representative rat model. ACKNOWLEDGMENTS The authors thank Arabe Ahmed from GlaxoSmithKline for the assistance in data gathering. They acknowledge Ir. Frank Lakiere and Dr. Andrei Postnov from the University of Antwerp for the technical assistance during the scanning of the animals. LITERATURE CITED Alfaro MF, Putney L, Tarkington BK, Hatch GE, Hyde DM, Schelegle ES. 2004. Effect of rapid shallow breathing on the distribution of 18O-labeled ozone reaction product in the respiratory tract of the rat. Inhal Toxicol 16:77–85. Andersen M, Sarangapani R, Gentry R, Clewell H, Covington T, Frederick CB. 2000. Application of a hybrid CFD-PBPK nasal do- 727 simetry model in an inhalation risk assessment: an example with acrylic acid. Toxicol Sci 57:312–325. Bush ML, Frederick CB, Kimbell JS, Ultman JS. 1998. A CFDPBPK hybrid model for simulating gas and vapor uptake in the rat nose. Toxicol Appl Pharmacol 150:133–145. Chapman RW, House A, Jones H, Richard J, Celly C, Prelusky D, Ting P, Hunter JC, Lamca J, Phillips JE. 2007. Effect of inhaled roﬂumilast on the prevention and resolution of allergen-induced late phase airﬂow obstruction in Brown Norway rats. Eur J Pharmacol 571:215–221. Chaturvedi A, Lee Z. 2005. Three-dimensional segmentation and skeletonization to build an airway tree data structure for small animals. Phys Med Biol 50:1405–1419. Coates MS, Chan HK, Fletcher DF, Chiou H. 2007. Inﬂuence of mouthpiece geometry on the aerosol delivery performance of a dry powder inhaler. Pharm Res 24:1450–1456. De Backer JW, Vos WG, Devolder A, Verhulst SL, Germonpre P, Wuyts FL, Parizel PM, De BW. 2007a. Computational ﬂuid dynamics can detect changes in airway resistance in asthmatics after acute bronchodilation. J Biomech 41:106–113. De Backer JW, Vos WG, Gorle CD, Germonpre P, Partoens B, Wuyts FL, Parizel PM, De BW. 2007b. Flow analyses in the lower airways: patientspeciﬁc model and boundary conditions. Med Eng Phys 30:872–879. Dong CC, Yin XJ, Ma JY, Millecchia L, Barger MW, Roberts JR, Zhang XD, Antonini JM, Ma JK. 2005. Exposure of brown Norway rats to diesel exhaust particles prior to ovalbumin (OVA) sensitization elicits IgE adjuvant activity but attenuates OVAinduced airway inﬂammation. Toxicol Sci 88:150–160. Ewing P, Eirefelt S, Andersson P, Blomgren A, Ryrfeldt A, Gerde P. 2008. Short inhalation exposures of the isolated and perfused rat lung to respirable dry particle aerosols; the detailed pharmacokinetics of budesonide, formoterol, and terbutaline. J Aerosol Med 21:169–180. Galle B, Ouyang H, Shi R, Nauman E. 2007. Correlations between tissue-level stresses and strains and cellular damage within the guinea pig spinal cord white matter. J Biomech 40:3029–3033. Greene E. 1970. Anatomy of the rat. New York: Macmillan Pub Co. Ji JH, Jung JH, Kim SS, Yoon JU, Park JD, Choi BS, Chung YH, Kwon IH, Jeong J, Han BS, Shin JH, Sung JH, Song KS, Yu IJ. 2007. Twenty-eight-day inhalation toxicity study of silver nanoparticles in Sprague–Dawley rats. Inhal Toxicol 19:857–871. Johnson KA. 2007. Imaging techniques for small animal imaging models of pulmonary disease: micro-CT. Toxicol Pathol 35:59–64. Kleinstreuer C, Shi H, Zhang Z. 2007. Computational analyses of a pressurized metered dose inhaler and a new drug-aerosol targeting methodology. J Aerosol Med 20:294–309. Kouadio K, Zheng KC, Tuekpe MK, Todoriki H, Ariizumi M. 2005. Airway inﬂammatory and immunological events in a rat model exposed to toluene diisocyanate. Food Chem Toxicol 43:1281–1288. Lam WW, Holdsworth DW, Du LY, Drangova M, McCormack DG, Santyr GE. 2007. Micro-CT imaging of rat lung ventilation using continuous image acquisition during xenon gas contrast enhancement. J Appl Physiol 103:1848–1856. Lee D, Fanucchi MV, Plopper CG, Fung J, Wexler AS. 2008. Pulmonary architecture in the conducting regions of six rats. Anat Rec 291:916–926. Lee MG, Wheelock AM, Boland B, Plopper CG. 2007. Long-term ozone exposure attenuates 1-nitronaphthalene induced cytotoxicity in nasal mucosa. Am J Respir Cell Mol Biol 38:300–309. Minard KR, Einstein DR, Jacob RE, Kabilan S, Kuprat AP, Timchalk CA, Trease LL, Corley RA. 2006. Application of magnetic resonance (MR) imaging for the development and validation of computational ﬂuid dynamic (CFD) models of the rat respiratory system. Inhal Toxicol 18:787–794. Schroeter JD, Kimbell JS, Bonner AM, Roberts KC, Andersen ME, Dorman DC. 2006. Incorporation of tissue reaction kinetics in a computational ﬂuid dynamics model for nasal extraction of inhaled hydrogen sulﬁde in rats. Toxicol Sci 90:198–207. WagnerU,StaatsP,FehmannHC,FischerA,WelteT,GronebergDA.2006. Analysis of airway secretions ina modelof sulfurdioxideinduced chronic obstructivepulmonarydisease(COPD).JOccupMedToxicol1:12. Wietholt C, Roerig DL, Gordon JB, Haworth ST, Molthen RC, Clough AV. 2008. Bronchial circulation angiogenesis in the rat quantiﬁed with SPECT and micro-CT. Eur J Nucl Med Mol Imaging 35:1124–1132.