We recruited participants from a local tertiary care center and through advertisement. Briefly,

Transcription

1 RSNA, /radiol Appendix E1 Materials and Methods Study Subjects We recruited participants from a local tertiary care center and through advertisement. Briefly, inclusion criteria comprised ambulatory men and women with a previous clinical diagnosis of COPD who were years old with a smoking history of 10 or more pack-years. Exclusion criteria included claustrophobia, current cigarette use, body mass index greater than 40, and contraindications to MR imaging (eg, metal, electronic, or magnetic implants). All participants were years old (mean age, 73 years ± 9; mean age in men, 73 years ± 9; age range in men, years; mean age in women, 71 years ± 7; age range in women, years; P =.4). In a single 2-hour visit, participants underwent the following procedures in the same order: (a) administration of salbutamol, (b) spirometry, (c) plethysmography and DL CO, (d) inspiration and expiration CT, and (e) MR imaging. Subjects with COPD were classified according to the Global Initiative for Chronic Obstructive Lung Disease (GOLD) (1). Pulmonary Function Postbronchodilator plethysmography and spirometry were performed with a body plethysmograph (MedGraphics, St. Paul, Minn) to measure the forced expiratory volume in 1 second (FEV 1 ), forced vital capacity (FVC), and static lung volumes, including total lung capacity, inspiratory capacity, residual volume, and functional residual capacity. The diffusing Page 1 of 9

2 capacity of the lung for carbon monoxide (DLCO) was also determined by using the attached gas analyzer. Measurements were performed according to the guidelines of the American Thoracic Society (2). MR Image Analysis 3 He static ventilation images were segmented by using a k-means approach that classifies voxel intensity values into five clusters ranging from signal void and hypointense to hyperintense, thereby generating a gas distribution cluster map. Delineation of the ventilation defect boundaries was performed by using a seeded region-growing algorithm that segmented the 1 H MR imaging thoracic cavity, a process that was previously described (3). 3 He MR imaging ADC values were determined by using a custom-built algorithm with MATLAB R2014b (MathWorks, Natick, Mass). To ensure that ADC values were generated for voxels that corresponded to ventilated lung regions, a k-means clustering algorithm was applied to the nondiffusion-weighted images to obtain a binary mask for each section (3). The resulting binary masks were applied to the corresponding nondiffusion-weighted images, and ADC maps were generated on a voxel-by-voxel basis according to Equation 1: S ADC b S 1 ln 0 (1), where S 0 is the segmented nondiffusion-weighted image, S is the diffusion-weighted image, and b = 1.6s/cm 2. CT Image analysis PRM measurements were generated by coregistering inspiratory and expiratory images with an affine method followed by a deformable step provided in the NiftyReg Page 2 of 9

3 ( package, which is ranked among the top registration algorithms for thoracic CT (4,5). The affine registration step was performed on a coarse-to-fine scheme to achieve both accuracy and computational efficiency. The following voxel-wise comparisons, in which coregistered inspiration and expiration images were imported into MATLAB and voxels were classified into four categories on the basis of expiration and inspiration thresholds, were performed, as was previously described: (a) for healthy tissue, inspiration greater than 950 HU and expiration greater than 856 HU; (b) for gas trapping, inspiration greater than 950 HU and expiration less than 856 HU; (c) for emphysema, inspiration less than 950 HU and expiration less than 856 HU; and (d) inspiration less than 950 HU and expiration greater than 856 HU (6). As was previously described (6), the fourth category of voxels was hypothesized to reflect noise in the data from registration error because, previously, these voxels did not correlate with FEV 1, FEV 1 /FVC, or the relative area of the CT attenuation histogram of less than 950 HU at inspiration. MR imaging to-ct Spatial Overlap 3 He MR imaging was registered to 1 H MR imaging by using a landmark-based approach, and 1 H MR imaging was registered to expiratory CT by using deformable registration by way of the modality-independent neighborhood descriptor (MIND) method, which previously was shown to be suitable for cross-modality image registration (3,7). The deformation field was applied to 3 He MR imaging sections for MR imaging to-ct coregistration. Deformable registration consisted of voxel-wise similarity measurements of MINDs of the two images rather than the images themselves, as well as diffusion regularization of the deformation field and optimization by using the Gauss-Newton framework. Registration was performed with three levels (ie, a down- Page 3 of 9

4 sampling factor of 4, 2, and 1) in a symmetric manner so the algorithm was not dependent on the choice of a moving or fixed image. Spatial overlap for 3 He ventilation defects and PRM voxels was evaluated by using the SOC, which is the intersection of 3 He ventilation defect and PRM voxels expressed as a fraction of the total number of CT PRM voxels or 3 He MR imaging ventilation defect voxels. The rationale for obtaining spatial overlap measurements in this manner was to determine the contribution of 3 He defect voxels within regions of PRM gas trapping or emphysema. The spatial overlap of PRM voxels within 3 He defect regions (MR imaging SOC) was also evaluated to determine the contribution of PRM (gas trapping and/or emphysema) to 3 He ventilation defects. CT and 3 He MR images were evaluated once because the CT analysis was fully automated, and the MR imaging semiautomated segmentation method was previously reported to provide high inter- and intraobserver reproducibility. In total, per subject, CT and MR image analysis was completed in approximately 1 hour (3). Statistics The variables considered were based on univariate relationships. The unstandardized coefficients were reported and show how a single-unit change in the independent variable influenced a change in the dependent variable. We also reported the standardized coefficients, which were expressed in units of standard deviation and showed the independent variable with the greatest coefficient, reflecting the greatest relative effect on the dependent variable. Multicolinearity was evaluated by using the variance inflation factor and was deemed acceptable when it was less than 10 (8). For spatial overlap coefficients, significant differences were measured Page 4 of 9

5 with paired t tests. Results were considered significant when the probability of two-tailed type I error was less than 5% (P <.05). Discussion Limitations The limitations of spirometry motivated us and others to continue to develop thoracic imaging methods, including ventilation MR imaging and CT PRM measurements to better phenotype cases of COPD by using direct and regional measurements of the underlying disease (9). Spirometry measurements of FEV 1 and FEV 1 /FVC are reproducible and inexpensive but provide only a global measure of lung function that is dominated by larger airway function. For this reason, pulmonary function tests are relatively insensitive to early disease stages (10). In addition, FEV 1 and FEV 1 /FVC are relatively poor surrogates for COPD symptoms and other outcomes, perhaps because disease heterogeneity derives from the pathologic features of COPD, including parenchyma destruction (ie, emphysema) and airway remodeling (ie, airways disease), which also differ in individual patients with the same FEV 1 (1,11,12). We recognize that this study has a number of limitations. First, this work was limited by the relatively small study group and the fact that we mainly evaluated ex-smokers with mildly abnormal and normal spirometry results. The study was prospectively planned and was driven by our interest in investigating very early or mild disease but, given our understanding of the heterogeneity of COPD patients, we must be cautious about extrapolating our results. We must also acknowledge that the results generated here were not compared with results stemming from commercially available software, such as Apollo Workstation 2.0 (VIDA Diagnostic, Coralville, Ia) or Lung Density Analysis software (IMBIO, Minneapolis, Minn). It should be noted that any Page 5 of 9

6 potential differences are likely to stem from the different image registration or warping algorithms used because, in general, the attenuation thresholds used are the same. Registration errors pose a challenge that was previously reported for PRM in the liver, and we caution that registration errors result in tissue misclassification because PRM analysis relies heavily on voxel-by-voxel comparisons (13). Notably, in the PRMs that were generated here, there was consistent scattering of misclassified voxels, which underscores the need for optimized thoracic CT registration techniques to minimize these effects. We recognize that MR imaging CT registration errors will certainly affect the analysis of spatial agreement. For this reason, we used a pulmonary MR imaging CT deformable registration method that was previously described as being highly suitable for cross-modality image registration and that achieved significantly better results than other methods, such as normalized mutual information (7). Differences between PRM and 3 He MR imaging may be related to the fact that PRM measurements are inherently more indirect, as they are based on the abnormal presence of air from both emphysema and airways disease. In contrast, inhaled gas methods provide static snapshots of regional ventilation. In advanced COPD, ventilation defects were previously shown to reflect both emphysema and airways disease, whereas in mild asthma, ventilation defects were shown to be directly related to abnormal airway wall thickening or airways disease (9,14). Similar to four-dimensional CT, Fourier decomposition MR imaging, and paired inspiratory and expiratory CT, PRM exploits the image signal differences from inspiration and expiration as air moves in and out and tissues contract and expand (15 19). All these approaches rely on either computational or intuitive coregistration of inspiratory and expiratory CT and assume that the abnormal presence of air can be regionally related to emphysema and/or functional small airways disease. Finally, hyperpolarized 3 He MR imaging is still limited to a few research facilities worldwide and is Page 6 of 9

7 unlikely to be translated clinically because of the depleted global supply of 3 He. However, with recent improvements in xenon-129 polarization and the development of fluorine-19 MR imaging, inhaled-gas MR imaging may yet be considered for regulatory approval and future clinical workflows (20,21). References 1. Vestbo J, Hurd SS, Agustí AG, et al. Global strategy for the diagnosis, management, and prevention of chronic obstructive pulmonary disease: GOLD executive summary. Am J Respir Crit Care Med 2013;187(4): Miller MR, Hankinson J, Brusasco V, et al. Standardisation of spirometry. Eur Respir J 2005;26(2): Kirby M, Heydarian M, Svenningsen S, et al. Hyperpolarized 3He magnetic resonance functional imaging semiautomated segmentation. Acad Radiol 2012;19(2): Modat M, McClelland J, Ourselin S. Lung registration using the NiftyReg package. Medical Image Analysis for the Clinic-A Grand Challenge. 2010; 2010: Murphy K, Van Ginneken B, Reinhardt JM, et al. Evaluation of registration methods on thoracic CT: the EMPIRE10 challenge. IEEE Trans Med Imaging 2011;30(11): Galbán CJ, Han MK, Boes JL, et al. Computed tomography-based biomarker provides unique signature for diagnosis of COPD phenotypes and disease progression. Nat Med 2012;18(11): Heinrich MP, Jenkinson M, Bhushan M, et al. MIND: modality independent neighbourhood descriptor for multi-modal deformable registration. Med Image Anal 2012;16(7): Page 7 of 9

8 8. O Brien RM. A caution regarding rules of thumb for variance inflation factors. Qual Quant 2007;41(5): Kirby M, Pike D, Coxson HO, McCormack DG, Parraga G. Hyperpolarized (3)He ventilation defects used to predict pulmonary exacerbations in mild to moderate chronic obstructive pulmonary disease. Radiology 2014;273(3): Vestbo J, Anderson W, Coxson HO, et al. Evaluation of COPD Longitudinally to Identify Predictive Surrogate End-points (ECLIPSE). Eur Respir J 2008;31(4): Agusti A, Calverley PM, Celli B, et al. Characterisation of COPD heterogeneity in the ECLIPSE cohort. Respir Res 2010;11: Pauwels RA, Buist AS, Calverley PM, Jenkins CR, Hurd SS; GOLD Scientific Committee. Global strategy for the diagnosis, management, and prevention of chronic obstructive pulmonary disease. NHLBI/WHO Global Initiative for Chronic Obstructive Lung Disease (GOLD) Workshop summary. Am J Respir Crit Care Med 2001;163(5): Lausch A, Chen J, Ward AD, Gaede S, Lee TY, Wong E. An augmented parametric response map with consideration of image registration error: towards guidance of locally adaptive radiotherapy. Phys Med Biol 2014;59(22): Svenningsen S, Kirby M, Starr D, et al. What are ventilation defects in asthma? Thorax 2014;69(1): Guerrero T, Sanders K, Castillo E, et al. Dynamic ventilation imaging from four-dimensional computed tomography. Phys Med Biol 2006;51(4): Page 8 of 9

9 16. Bauman G, Puderbach M, Deimling M, et al. Non-contrast-enhanced perfusion and ventilation assessment of the human lung by means of fourier decomposition in proton MRI. Magn Reson Med 2009;62(3): Bommart S, Marin G, Bourdin A, et al. Relationship between CT air trapping criteria and lung function in small airway impairment quantification. BMC Pulm Med 2014;14: Hersh CP, Washko GR, Estépar RS, et al. Paired inspiratory-expiratory chest CT scans to assess for small airways disease in COPD. Respir Res 2013;14: Kim EY, Seo JB, Lee HJ, et al. Detailed analysis of the density change on chest CT of COPD using non-rigid registration of inspiration/expiration CT scans. Eur Radiol 2015;25(2): Mugler JP 3rd, Altes TA. Hyperpolarized 129Xe MRI of the human lung. J Magn Reson Imaging 2013;37(2): Ruiz-Cabello J, Barnett BP, Bottomley PA, Bulte JW. Fluorine (19F) MRS and MRI in biomedicine. NMR Biomed 2011;24(2): Page 9 of 9