| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Thoracic Imaging |
1 From the Laboratoire de Physiologie Cellulaire Respiratoire (Institut National de la Santé et de la Recherche Médicale E-0356), Université Victor Ségalen, Bordeaux, France (P.B., J.M.T.d.L., R.M., F.L.); and Unité d’Imagerie Thoracique et Cardiovasculaire, Hôpital Cardiologique Haut-Lévêque, Avenue de Magellan, 33604 Pessac, France (V.P., P.D., F.L.). Received January 26, 2004; revision requested April 2; final revision received August 12; accepted September 15. Supported by grants from Programme Hospitalier de Recherche Clinique in 1997 and 2002. Address correspondence to F.L. (e-mail: francois.laurent@chu-bordeaux.fr).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: All subjects gave written informed consent. The study was approved by local ethics committee. With Laplacian of Gaussian algorithm, software was tested in phantom and excised sheep lung fixed in inflation and validated with Bland-Altman analysis. Study prospectively included nine nonsmokers (six women, three men; mean age, 53 years ± 5.6 [standard error of the mean]) with normal lung function (group 1), seven smokers (three women, four men; mean age, 56 years ± 5.6) with normal lung function (group 2), and eight smokers (zero women, eight men; mean age, 65 years ± 4.0) with COPD. Calculations were determined with spirometrically gated CT: For each selected bronchus, the wall area (WA), internal area (IA), airway caliber (sum of IA and WA), and WA/IA ratio were calculated. For each patient, summation of WA to summation of IA (
WA/IA) ratio, which reflected normalized airway wall thickness, was calculated. Groups were compared by using analysis of variance with generalized linear model and unpaired t test. Pearson correlation coefficient was used to assess correlation between software measurements and pulmonary function test results.
RESULTS: Comparison of measurements in phantom and excised sheep lung with algorithm measurements revealed that the latter were reliable and repeatable. In clinical study, WA/IA ratio was significantly different among three groups (P < .001). Normalized airway wall thickness and IA were significantly related to lung function test data, including forced expiratory volume in 1 second (r = –0.54, P = .006), specific airway conductance (r = –0.45, P = .03), and forced expiratory flow between 25% and 75% of vital capacity (r = –0.65, P < .001).
CONCLUSION: This software provides accurate and reproducible measurements of IA and WA of bronchi on thin-section CT images and demonstrates that in vivo normalized airway wall thickness was larger in smokers with COPD than it was in smokers or nonsmokers without COPD.
© RSNA, 2005
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
Thin-section computed tomography (CT) is increasingly used for the study of airway morphologic characteristics in vivo. Theoretically, thin-section CT can depict the dimensions of airways as narrow as approximately 1–2 mm in diameter. For clinical purposes, the use of thin-section CT for assessment of bronchial wall thickness in patients with airway diseases has been, so far, mainly subjective (4). Although subjective grading has led to new insights into the pathophysiologic features of various airway diseases, objective methods are suitable to perform longitudinal studies and to compare airway dimensions before and after bronchial provocation testing or therapeutic intervention in chronic diseases (5). Therefore, it has been suggested that CT be used to evaluate airway dimensions in patients with COPD (6), and quantitative algorithms for analysis of thin-section CT images have been designed and validated in studies with phantoms and animals (7). Although a large number of studies have been conducted to detect and to assess emphysema, for example, by using CT, very few have been devoted to measurement of airway dimensions in patients with COPD. Nakano and colleagues (8) evaluated the airway wall dimension at the origin of the apical bronchus of the upper lobe of the right lung in smokers. Since a single bronchus was measured in each patient, however, the heterogeneity of inflammation within the bronchial tree was not assessed (8). Little et al (9) measured airway caliber (sum of internal area [IA] and wall area [WA], hereafter referred to as IA + WA) in a variety of different bronchi in asthmatic patients, but their method was manual and, thus, observer dependent. IA is defined by the area limited by the internal border of a bronchus wall.
The aim of the present study, therefore, was to design and validate a dedicated software tool to measure airway dimensions on thin-section CT images and to use the tool to prospectively compare airway wall thickness in nonsmokers with normal lung function with that in smokers with and without COPD.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
Image Analysis Software: DACLOG
A Laplacian of Gaussian algorithm was used to develop a software tool for extraction of bronchial geometric parameters defined as IA and WA from thin-section CT images. Laplacian filters are second-derivative filters used to find areas of rapid change (edges) on images. Since they are very sensitive to noise, it is common to smooth the image (eg, by using a Gaussian filter) before application of the Laplacian filter, a one-step process called the Laplacian of Gaussian operation (10) (Fig 1). The software was implemented on a personal computer (Maxdata, Würselen, Germany). The first step for the image analysis was to import thin-section CT images by means of a local area network in Digital Imaging and Communications in Medicine format. With the software, we were able to assess the IA and WA of any bronchus visible on the thin-section CT image. Although there was no limitation in regard to the number of bronchi studied, the software was developed to enable evaluation of five bronchi on the same section. After the observer (V.P. or F.L.) independently traced a square region of interest that encompassed the selected bronchus and, therefore, that depended on the size of the selected bronchus, the algorithm was used to calculate a discrete 3 x 3 Laplacian of Gaussian mask. A spatial convolution was applied to the whole image to get an accurate depiction of airway contours. The selected region of interest was then displayed and magnified with a factor of five to make the region more visible to the operator. The resultant image was binary: white pixels represented the bronchial wall and black pixels represented the bronchial IA. When this first step did not allow complete separation of the bronchial wall from adjacent structures such as blood vessels or when the bronchus contour was not closed, manual editing of pixels was performed by two authors (V.P. or F.L.) independently (Fig 1, E–G). IA and WA were then calculated by using a growing algorithm. A white pixel was selected within the bronchial wall, and all adjacent pixels connected at four points were aggregated automatically as far as the edge of the wall; the same operation was repeated for the IA by using a black pixel. Last, the number of pixels was converted to an area expressed in square millimeters.
|
Excised sheep lung study.—A 1.350-kg lung was removed from a freshly sacrificed sheep in the local slaughterhouse and was fixed in inflation with formalin fumes by using a method adapted from the technique described by Weibel and Vidone (11). Once the lung was fixed, it was cut into 1-cm-thick transverse slices perpendicular to its largest axis, in a plane similar to that which is used with CT.
Airways on the cut surface of each thick slice were visualized and photographed by using a method adapted from King and colleagues (7). This technique was thought to be the most suitable to match the thin-section CT images with the digitized photographs of the slices. The images were digitized with a high-resolution camera (Nikon, Tokyo, Japan). Two rulers with a graduation of 0.5 mm were included on the image for calibration. The rulers were positioned at 90° to minimize the measurement errors caused by image distortions. The images were transferred to a personal computer. The measurement of airways was performed manually with graphic analysis software (Scion Image, version beta 4.0.2; Scion, Frederick, Md). Care was taken to perform a calibration before each measurement to prevent image distortion. The software measured the area within the external border of the bronchus (or EA) and IA; WA was calculated (WA = EA – IA). After the selection of bronchi to be analyzed by one author (V.P.), two independent observers (V.P., F.L.) performed the measurements of each selected bronchi. The first observer (V.P.) performed two measurements.
The lung slices were then scanned by using thin-section CT. To ensure the best fit between thin-section CT images and digitized images, each thick slice of the lung was placed between two cardboard sheets, which were oriented in the scanning plane to ensure that as much of the slice surface was in the acquisition plane of the scanner as possible. The volume of acquisition encompassed the whole lung volume between cardboard sheets for each thick slice. Therefore, because the lung surface of the thick slice was not perfectly flat, the next contiguous 1-mm slice was used for analysis if parts of the cut surface were incomplete on the first 1-mm slice. The same scanner, settings, and reconstruction algorithm that were used for the images of the phantom were used for those of the excised sheep lung. Selected thin-section CT images of the cut surface were segmented by using DACLOG, and the same two independent observers performed the measurements of each selected bronchus.
Clinical Study
Subjects.—The study prospectively included 24 subjects during 2 months, without any occupational dust exposure, as follows: nine nonsmokers (mean age, 53 years ± 5.6 [standard error of the mean]; six women and three men) who never smoked and had normal lung function (group 1); seven smokers (mean age, 56 years ± 5.6; three women and four men) with normal lung function (group 2); and eight smokers (mean age, 65 years ± 4.0; zero women and eight men) with COPD (group 3). COPD was assessed by one author (P.B.) and was defined by a marked decrease in forced expiratory volume in 1 second (FEV1), specific airway conductance, and forced expiratory flow between 25% and 75% of vital capacity (FEF25%–75%). All subjects gave their written informed consent to participate in the study, which was approved by the local ethics committee. The time between performance of thin-section CT and that of pulmonary function tests was less than 7 days.
Thin-section CT and analysis.—Single-section thin-section CT scans were obtained with the same CT unit as was used for the phantom study in sequential and spiral modes for assessment of emphysema and airway dimensions, respectively. The spiral mode was chosen for assessment of airway dimensions to scan a large volume of data during a breath hold, despite the use of single-section CT. Sequential 1-mm-thick thin-section CT sections were acquired every 10 mm (120-kV voltage, 70-mA current, 1-mm collimation, and 750-msec rotation time) and encompassed both lungs, were reconstructed on a 35-cm field of view with a 512 x 512 matrix, and were visualized on lung windows (window width = 1800 HU, window level = –700 HU). The visual score described by Goddard et al (12) was used to obtain a qualitative CT score for emphysema. Destruction of a certain percentage of the lung caused by emphysema was identified by one author (F.L., with 15 years of experience), with the following scale: score of 1, destruction of 1%–25%; score of 2, destruction of 26%–50%; score of 3, destruction of 51%–75%; and score of 4, destruction of more than 75% of the lung. The sum of the scores for each of the three lobes was used as the global severity score. The maximum possible score was 12 per lung.
In addition, for measurement of airway dimensions, single-volume thin-section CT scans were acquired in the spiral mode with parameters similar to those used for the excised sheep lung study (1-mm collimation, 750-msec rotation time, pitch of 2), through an anatomic region of interest of 40-mm length starting 4 cm below the level of the carina in the lower lobe of the right lung. We used spirometric gating at a level of breath hold corresponding to 90% of vital capacity (13). This method was used to trigger signals for scans to be obtained at a selected level of respiration and to interrupt airflow during scanning. The level of inspiration was defined by means of a small hand-held transducer (Micro Medical Instruments, Rochester, England) through which the patient was asked to breathe. A microcomputer that was connected to the transducer was used to determine vital capacity and to generate trigger signals for scans to be obtained at a user-selected level of inspiration. These levels could be chosen in percentage of vital capacity. As the trigger signal was generated and sent to the CT scanner, airflow was interrupted mechanically by the closing of a valve attached to the transducer. Therefore, the momentary status was kept constant for the duration of CT scanning.
All the patients were alert and cooperative and prepared for the examination by practicing the breathing maneuver before the CT study. When the patient was positioned in the scanner, a spirometric measurement of vital capacity was obtained. Then, acquisitions were performed at 90% of vital capacity. The patients completed the study within 10 minutes. Reconstructions were obtained by one author (F.L., with 15 years of experience) with a 13-cm field of view and a 512 x 512 matrix. These sections were then used for image evaluation and were transferred to a workstation for image analysis with DACLOG.
The area of the lung sampled for the study was situated in the lower lobe of the right lung to avoid as many cardiac artifacts as possible, and all sections that were free from artifacts were used. According to the results of our phantom study, measurements were performed on all bronchi with a ratio of largest bronchial diameter to smallest bronchial diameter of less than 1.50. A mean number of 25 sections were available per patient. For the purpose of a comparison assessed by one author (P.B.), in addition to WA and IA, the IA + WA value and the WA/IA ratio for each selected bronchus were calculated. For each patient, the summation of WA to the summation of IA (ie, WA/IA) ratio and the summation of WA to the summation of IA + WA (ie, WA/[IA + WA]) ratio were also determined. To evaluate the relative contribution of airway size, images of the bronchi were subcategorized in four sets according to the IA + WA value (set A, <10 mm2; set B, between 10 and 15 mm2; set C, between >15 and 20 mm2; and set D, >20 mm2). The lung attenuation in the parenchyma surrounding the analyzed bronchi was also measured on the sections selected for airway dimension measurements by the observer (F.L.) with use of round regions of interest of 10–15 mm in diameter in areas away from blood vessels and equally distributed between anterior (nondependent), lateral, and posterior (dependent) areas (14). Mean values of lung attenuation were compared.
Statistical Analysis
Since WA and IA were not normally distributed, values were log transformed. In the phantom study, the accuracy of measurements with DACLOG was compared with that of measurements with the micrometer and with data obtained at various angles on the segmented image by using Pearson correlation coefficients and one-tailed paired t tests. According to Bland-Altman analysis (15–17), the reliability of DACLOG was evaluated by one author (P.B.) after log transformation of data to obtain more normal data by using (a) the Pearson correlation coefficient, (b) the lack of agreement (ie, bias estimated with the mean difference and the standard deviation of the difference), (c) the intraclass correlation coefficient, and (d) the means of DACLOG values and those of the observer plotted against their differences. The repeatability of DACLOG was also analyzed by the same author as follows: (a) The measurement error was evaluated graphically, with the plotting of the individual subject standard deviations against their means, and analytically, with the Spearman rank correlation coefficient. (b) The within-subject standard deviation also was evaluated. For the clinical study, comparison between groups of patients and/or sets of airways was determined by the same author by using the following: (a) one- and two-way analysis of variance (ANOVA); (b) ANOVAs with generalized linear model procedures, with consideration of both between- and within-subject variation; and (c) unpaired t test of DACLOG measurements and the Pearson correlation coefficient for determination of the correlation between the DACLOG measurement and the pulmonary function test results. Results were considered significant when there was a difference with P < .05. All analyses were performed with software (NCSS 2001; NCSS Statistical Software, Kaysville, Utah).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
We took into account both the various sizes of the bronchi and the heterogeneity of airway thickness by using the WA/IA ratio (Fig 6e) or the WA/ (IA + WA) ratio (Fig 6f) for each patient. Since the number of bronchi per patient varied, values for ratios were adjusted accordingly. Both WA/IA and WA/ (IA + WA) ratios (F = 7.81, P = .003 and F = 7.52, P = .004, respectively, n = 24) were significantly different within the three groups, with generalized linear model ANOVA and two degrees of freedom. These two ratios differed significantly between healthy control subjects (group 1) and smokers without COPD (group 2, Fig 6e, 6f). In addition, values for functional parameters that reflected airway obstruction (ie, FEV1, specific airway conductance, FEF25%–75%) were significantly correlated with the IA value, the WA/IA ratio, or the WA/IA ratio (Table 4). Representative images from subjects in the three groups are illustrated in Figure 7.
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
WA/IA) that can be used to discriminate between smokers and nonsmokers even in the absence of COPD. To circumvent interobserver variability and parallax errors ascribed to manual methods (6), we used dedicated software for tracing internal and external contours of bronchi. The resulting contour was also independent from the thin-section CT window setting, a fundamental condition for assessment of airway WA (18,19). The method is semiautomatic and allows correction of the airway external wall contour despite its connection to a blood vessel, without assumption of perfect roundness and symmetry of the WA. Previous investigators developed algorithms that allow measurement of airway lumen and WAs with CT.
Okazawa et al (20) and McNamara and associates (21) described a technique in which the internal and external perimeters of the airways are traced on images obtained at a window width of 1500 HU and at a window level of –450 HU. Amirav et al (19) developed a computerized algorithm for measurement of airway lumen area that was based on an edge detection contour algorithm by using the full-width-at-half-maximum principle, which has less subjectivity and greater speed. Webb et al (22) and Brown and colleagues (23) used algorithms for which the operator draws radial lines from the airway lumen through the whole thickness of the airway wall.
Wood et al (24) developed a threshold method to define airway lumen area; the lumen centroid was used to define the central axis of the airway so that the angle of orientation could be measured. Subsequently, these authors used this algorithm to perform a three-dimensional reconstruction of the airway tree after they converted the data into isotropic voxels and performed accurate measurements of cross-sectional diameters of phantom tubes at various angles (24). King et al (7) developed and validated an automatic image analysis algorithm with which they were able to take into account the airway angulations by using only three contiguous sections and to correct the effect of airway orientations on the values of luminal areas and WAs.
Prêteux et al (25) developed an automatic method of segmentation that was based on mathematical morphology (image processing tool for measurement of morphologic features) and that was able to provide an accurate cross section of airways larger than 4 mm in diameter. Owing to the anatomy of the lung, most of the airways likely run oblique to the plane of the section rather than perpendicular to it. Therefore, on a two-dimensional thin-section CT image, there is an error in the calculation of IA and WA, depending on how acutely the airway is angled with relation to the acquisition plane. In our phantom study, this error was acceptable when the analysis was restricted to the airways with a largest bronchial diameter–to–smallest bronchial diameter ratio that was smaller than 1.50.
Discrepancies between thin-section CT measurements and those performed on specimens can be related to volume averaging of the folds along the mucosal surface and to irregularities on the adventitial surface of the airway at sites of parenchymal attachment in obliquely orientated airways. In addition, CT scanners also have a point-spread function in which the attenuation value of any pixel is affected by the attenuation value of the adjacent pixels. As a consequence, this has the effect of enlarging small tubular structures that run parallel to the CT axis and causes artifactual thickening of thin planar structures such as airway walls.
Therefore, validation of such dedicated software before its use for clinical purposes is required. For validation, we used fixed inflated excised sheep lungs. With excised sheep lungs, the effect of volume averaging can be taken into account, whereas this cannot be done with an airway phantom, in which irregularities of the mucosal and adventitial surfaces and presence of adjacent blood vessels cannot be duplicated. The techniques that have been published have been validated by using data from phantom studies (7,19,21,24) and excised animal lungs (7) or by developing a realistic modeling of airways included in CT scans of animal lungs obtained in vivo (25).
Since a history of smoking is associated with bronchial inflammation (1,26,27), even in the absence of COPD (3), we evaluated airway dimensions with thin-section CT in three groups of patients; we compared data in smokers with COPD and smokers without COPD with the data in healthy nonsmoking control subjects. We found that the WA was increased in smokers with COPD compared with that in healthy control subjects. This parameter, however, was not different between healthy control subjects and smokers without COPD or between smokers with COPD and smokers without COPD.
In vivo, WA seems to be more suitable, in theory, than IA for assessment of pathologic alterations in bronchi, because IA could be affected by both bronchial thickness and loss of elastic recoil caused by emphysema (14). In our study, we found that IA was related to functional parameters (ie, FEV1, FEF25%–75%), whereas WA was not. In addition, IA was significantly different in smokers with COPD and in those without COPD. Since bronchial pathologic findings are not uniformly distributed throughout the whole bronchial tree, we searched for parameters that reflect both heterogeneity and various sizes of bronchi. The WA/IA ratio was mostly affected by IA and could not be used to discriminate healthy control subjects from smokers without COPD. In contrast, the WA/IA ratio, which is calculated by using various bronchi in each patient, was significantly different among the three groups.
To the best of our knowledge, in only one study have the researchers evaluated airway dimensions with CT in smokers who had COPD (8). The authors showed that not only IA but also WA, assessed by using manual CT analysis of a single bronchus (ie, apical bronchus of the upper lobe of the right lung), was correlated, although weakly, with a decrease in FEV1. The absence of correlation between WA and functional parameters in our study could be related to our small sample size (n = 24). The discrepancy between our results and those obtained by Nakano et al (8) could be explained by the absence of a control group in the latter study. Although our sample size was limited, we analyzed a large range of bronchi per patient and compared airway parameters with those in a control group. We also took into account the potential heterogeneity of pathologic findings with evaluation of various bronchus sizes.
As did Nakano et al (8), we evaluated global lung emphysema. Moreover, we also analyzed emphysema in the lung surrounding the bronchi that we examined. We found that, whereas patients with COPD have a greater amount of emphysema determined by means of the thin-section CT score for emphysema, especially in the upper lobes, the mean lung attenuation was similar within the surrounding parenchyma of the examined bronchi among the three groups. In group 3 of this study, the score for emphysema was lower than was usually found in patients with COPD disease. Nevertheless, our goal in this preliminary study was only to verify the absence of the role of emphysema in determination of the bronchial diameter, and this role clearly needs further investigation to evaluate how emphysema and bronchial abnormalities in such disease affect the bronchial diameter. CT sensitivity for detection of emphysema, however, is lower than sensitivity for detection at pathologic analysis (28). Therefore, we cannot exclude a loss of elastic recoil to explain the lower airway IA in patients with COPD. Further improvements, which include multisection CT acquisition, should enable us to analyze bronchial thickness of the whole bronchial tree in a procedure with a single acquisition and, thus, to optimize the measurement of the WA/IA ratio parameter.
Our study had several limitations. The first one was the small number of subjects included, but it was a validation study, and further studies need to be performed to determine the clinical importance of these results. Therefore, this study was also affected by imperfect matching in terms of sex, age, and history of smoking among groups, and this may have affected the bronchial thickness, since a correlation of the bronchoarterial ratio with age and smoking has been found in asymptomatic subjects (29). Another limitation was that we chose the bronchi of the lower lobe of the right lung for assessment of the bronchial wall thickness in humans. This site was chosen because it allowed a more convenient method to obtain a tangential view of the bronchus and artery; therefore, the number of pixels that needed to be edited from the automated segmentation was limited. In addition, transmitted cardiac motion artifacts obscure the lower lobe of the right lung less than they do that of the left lung. Further studies are needed to evaluate the technique for other parts of the lung, such as the upper lobes, but a multisection acquisition will allow the whole lung to be scanned.
In conclusion, we validated an algorithm that can be used clinically to measure airway lumen and WA on thin-section CT images and determined a new parameter that reflects airway wall thickness and that allows discrimination of healthy subjects from smokers with COPD or smokers without COPD. By using these measurements, we found larger normalized wall thickness in smokers with airway obstruction.
|
FOOTNOTES |
|---|
Authors stated no financial relationship to disclose.
Author contributions: Guarantors of integrity of entire study, F.L., P.B.; study concepts, F.L., J.M.T.d.L., P.B.; study design, F.L., P.B., J.M.T.d.L.; literature research, V.P., P.B., F.L.; clinical studies, V.P., P.D.; experimental studies, P.B., V.P., P.D.; data acquisition, V.P., P.B., F.L.; data analysis/interpretation, P.B., F.L., J.M.T.d.L., R.M.; statistical analysis, P.B., R.M.; manuscript preparation and editing, P.B., F.L.; manuscript definition of intellectual content and revision/review, P.B., F.L., J.M.T.d.L., R.M.; manuscript final version approval, all authors
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  M. Cazzola, W. MacNee, F. J. Martinez, K. F. Rabe, L. G. Franciosi, P. J. Barnes, V. Brusasco, P. S. Burge, P. M. A. Calverley, B. R. Celli, et al. Outcomes for COPD pharmacological trials: from lung function to biomarkers Eur. Respir. J., February 1, 2008; 31(2): 416 - 469. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Zhang, J. G. Fletcher, T. J. Vrtiska, A. Manduca, J. L. Thompson, M. L. Raghavan, R. J. Wentz, and C. H. McCollough Large-Vessel Distensibility Measurement with Electrocardiographically Gated Multidetector CT: Phantom Study and Initial Experience Radiology, October 1, 2007; 245(1): 258 - 266. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Montaudon, P. Berger, G. de Dietrich, A. Braquelaire, R. Marthan, J. M. Tunon-de-Lara, and F. Laurent Assessment of Airways with Three-dimensional Quantitative Thin-Section CT: In Vitro and in Vivo Validation Radiology, December 19, 2006; (2006) 2422060029. [Abstract] [Full Text]
|
|
|
|
|
|
M. Montaudon, P. Berger, A. Cangini-Sacher, G. de Dietrich, J. M. Tunon-de-Lara, R. Marthan, and F. Laurent Bronchial Measurement with Three-dimensional Quantitative Thin-Section CT in Patients with Cystic Fibrosis Radiology, December 19, 2006; (2006) 2422060030. |
