Pulmonary emphysema subtypes defined by unsupervised machine learning on CT scans

Elsa D Angelini, Jie Yang, Pallavi P Balte, Eric A Hoffman, Ani W Manichaikul, Yifei Sun, Wei Shen, John H M Austin, Norrina B Allen, Eugene R Bleecker, Russell Bowler, Michael H Cho, Christopher S Cooper, David Couper, Mark T Dransfield, Christine Kim Garcia, MeiLan K Han, Nadia N Hansel, Emlyn Hughes, David R Jacobs, Silva Kasela, Joel Daniel Kaufman, John Shinn Kim, Tuuli Lappalainen, Joao Lima, Daniel Malinsky, Fernando J Martinez, Elizabeth C Oelsner, Victor E Ortega, Robert Paine, Wendy Post, Tess D Pottinger, Martin R Prince, Stephen S Rich, Edwin K Silverman, Benjamin M Smith, Andrew J Swift, Karol E Watson, Prescott G Woodruff, Andrew F Laine, R Graham Barr, Elsa D Angelini, Jie Yang, Pallavi P Balte, Eric A Hoffman, Ani W Manichaikul, Yifei Sun, Wei Shen, John H M Austin, Norrina B Allen, Eugene R Bleecker, Russell Bowler, Michael H Cho, Christopher S Cooper, David Couper, Mark T Dransfield, Christine Kim Garcia, MeiLan K Han, Nadia N Hansel, Emlyn Hughes, David R Jacobs, Silva Kasela, Joel Daniel Kaufman, John Shinn Kim, Tuuli Lappalainen, Joao Lima, Daniel Malinsky, Fernando J Martinez, Elizabeth C Oelsner, Victor E Ortega, Robert Paine, Wendy Post, Tess D Pottinger, Martin R Prince, Stephen S Rich, Edwin K Silverman, Benjamin M Smith, Andrew J Swift, Karol E Watson, Prescott G Woodruff, Andrew F Laine, R Graham Barr

Abstract

Background: Treatment and preventative advances for chronic obstructive pulmonary disease (COPD) have been slow due, in part, to limited subphenotypes. We tested if unsupervised machine learning on CT images would discover CT emphysema subtypes with distinct characteristics, prognoses and genetic associations.

Methods: New CT emphysema subtypes were identified by unsupervised machine learning on only the texture and location of emphysematous regions on CT scans from 2853 participants in the Subpopulations and Intermediate Outcome Measures in COPD Study (SPIROMICS), a COPD case-control study, followed by data reduction. Subtypes were compared with symptoms and physiology among 2949 participants in the population-based Multi-Ethnic Study of Atherosclerosis (MESA) Lung Study and with prognosis among 6658 MESA participants. Associations with genome-wide single-nucleotide-polymorphisms were examined.

Results: The algorithm discovered six reproducible (interlearner intraclass correlation coefficient, 0.91-1.00) CT emphysema subtypes. The most common subtype in SPIROMICS, the combined bronchitis-apical subtype, was associated with chronic bronchitis, accelerated lung function decline, hospitalisations, deaths, incident airflow limitation and a gene variant near DRD1, which is implicated in mucin hypersecretion (p=1.1 ×10-8). The second, the diffuse subtype was associated with lower weight, respiratory hospitalisations and deaths, and incident airflow limitation. The third was associated with age only. The fourth and fifth visually resembled combined pulmonary fibrosis emphysema and had distinct symptoms, physiology, prognosis and genetic associations. The sixth visually resembled vanishing lung syndrome.

Conclusion: Large-scale unsupervised machine learning on CT scans defined six reproducible, familiar CT emphysema subtypes that suggest paths to specific diagnosis and personalised therapies in COPD and pre-COPD.

Keywords: COPD epidemiology; Emphysema; Imaging/CT MRI etc.

Conflict of interest statement

Competing interests: EDA, PPB, AM, YS, WS, JHMA, MHC, DC, EH, DRJ, SK, JDK, TL, JL, ECO, WP, MRP, SSR, EKS, KEW and AFL reports receiving grants from the National Institutes of Health (NIH). JY performed the work at Columbia University but is now an employee of Google. EAH reports receiving grants from the NIH; being a founder and shareholder of VIDA Diagnostics; and holding patents for an apparatus for analysing CT images to determine the presence of pulmonary tissue pathology, an apparatus for image display and analysis, and a method for multiscale meshing of branching biological structures. EBA reports receiving grants from the American Heart Association and the NIH. CBC reports receiving personal fees from GlaxoSmithKline. MTD reports receiving a grant from the NHLBI and personal fees from AstraZeneca, GlaxoSmithKline, Pulmonx, PneumRx/BTG and Quark. MKH reports consulting for GlaxoSmithKline, AstraZeneca and Boehringer Ingelheim receiving research support from Novartis and Sunovion. NNH reports receiving grants from the NIH, Boehringer Ingelheim, and the COPD Foundation. JDK reports receiving grants from US Environmental Protection Agency and the NIH. FJM reports serving on COPD advisory boards for AstraZeneca, Boehringer Ingelheim, Chiesi, GlaxoSmithKline, Sunovion and Teva; serving as a consultant for ProterixBio and Verona; serving on the steering committees of studies sponsored by the NHLBI, AstraZeneca, and GlaxoSmithKline; having served on data safety and monitoring boards of COPD studies supported by Genentech and GlaxoSmithKline. BMS reports receiving grants from the NIH, Canadian Institutes of Health Research (CIHR), Fonds de la recherche en santé du Québec (FRQS), the Research Institute of the McGill University Health Centre, the Quebec Lung Association and AstraZeneca. PGW reports receiving personal fees for consultancy from Theravance, AstraZeneca, Regeneron, Sanofi, Genentech, Roche and Janssen. RGB reports receiving grants from the COPD Foundation, the US Environmental Protection Agency (EPA), the American Lung Association and the NIH.

© Author(s) (or their employer(s)) 2023. No commercial re-use. See rights and permissions. Published by BMJ.

Figures

Figure 1.
Figure 1.
Schema of unsupervised machine learning, data reduction, primary descriptive analyses, events analyses and GWAS. Unsupervised machine learning of possible emphysema subtypes was performed in twoindependent training sets in SPIROMICS. Both training sets yielded 10 possible emphysema subtypes, and training was repeated on all of SPIROMICS. The resultant 10 possible emphysema subtypes were labelled on MESA Lung CT scans. Data reduction was performed in SPIROMCS and MESA Lung and yielded six CT emphysema subtypes; data reduction was confirmed longitudinally on coregistered CT scans in a subset of the MESA Lung Study oversampled for COPD and smoking. Primary descriptive analyses of these subtypes were performed in the MESA Lung Study. Cardiac scans in MESA were labelled for the Event Analyses in MESA. GWAS Discovery was performed in SPIROMICS; replication of genetic results occurred on labelled cardiac scans in MESA and MESA SHARe and in COPDGene.
Figure 2.
Figure 2.
Representative visual illustrations of the six CT emphysema subtypes. Coronal views of lungs on CT scans and the corresponding labelled masks with the discovered CT emphysema subtypes on predominantly affected sample cases (i.e. with proportion of a certain CT emphysema subtype being much larger than any other). Color coding of CT emphysema subtypes is the same across examples; grey labelling denotes non-emphysematous regions. Abbreviation: CPFE=combined pulmonary fibrosis/emphysema
Figure 3.
Figure 3.
Distributions of the six discovered CT emphysema subtypes in SPIROMICS and the MESA Lung Study. Mean percentages of CT emphysema subtypes in SPIROMICS, a COPD case-control study of 2655 participants with 20 or more packyears of smoking (median packyears 43.0; 66.2% with COPD) and 198 non-smoking controls, and in the MESA Lung Study, a population-based study of 2,949 participants, 54.2% of whom had ever smoked cigarettes (median packyears 14.5) and 16.9% with COPD. Abbreviations: CPFE=combined pulmonary fibrosis/emphysema, COPD=chronic obstructive pulmonary disease
Figure 4:
Figure 4:
Multivariable associations of CT emphysema subtypes with symptoms, physiology, lung structure, and lung function decline in the MESA Lung Study.* β estimates for continuous outcomes show the effect size per 10% increment in CT emphysema subtype, except for percent emphysema, which is per 1% increment in CT emphysema subtype. The β estimates for chronic bronchitis and interstitial lung abnormalities are the log(odds ratios). All results adjusted for age, sex, race/ethnicity, height, weight, smoking status, pack-years, scanner manufacturer and other CT emphysema subtypes. Abbreviations: CPFE=combined pulmonary fibrosis/emphysema; FEV1= Forced expiratory volume in one second; FVC=Forced expiratory volume in one second.
Figure 5.
Figure 5.
Manhattan and local association plots for the three genome-wide significant, replicated gene variants for three CT emphysema subtypes in SPIROMICS. The red lines show the level of statistical significance (P = 5×10–8). The genome-wide significant SNP for the Combined Bronchitis-Apical Emphysema subtype replicated among Whites (P=0.01) and the entire replication sample (P=0.04). The genome-wide significant SNP for the restrictive CPFE subtype replicated among Whites (P=0.01) and the entire replication sample (P=0.04). The first genome-wide significant SNP for the obstructive CPFE subtype on chromosome 19 had variance only among Black participants and replicated in this sample (P=0.046). The second genome-wide significant SNP for the obstructive CPFE subtype on chromosome 16 did not replicate. There were no significant replicated genetic associations for the diffuse and senile CT emphysema subtypes (not shown).* Results are shown for the lowest attenuation (most severe) of the three preliminary subtypes that comprise the Combined Bronchitis-Apical Emphysema subtype. Abbreviation: CPFE=combined pulmonary fibrosis/emphysema

References

    1. World Health Organization. The top 10 causes of death, 2019. Geneva, Switzerland: WHO; 2020 [Available from: accessed 7-7-21.
    1. Shrine N, Guyatt AL, Erzurumluoglu AM, et al. New genetic signals for lung function highlight pathways and chronic obstructive pulmonary disease associations across multiple ancestries. Nat Genet 2019;51(3):481–93.
    1. Global Stategy for the Diagnosis, Management, and Prevention of Chronic Obstructive Pulmonary Disease - 2023 Report: Global Initiative for Chronic Obstructive Lung Disease, 2022.
    1. Baldwin ED, Cournand A, Richards DW Jr. Pulmonary insufficiency; a study of 122 cases of chronic pulmonary emphysema. Medicine 1949;28:201–37.
    1. Burrows B, Fletcher CM, Heard BE, et al. The emphysematous and bronchial types of chronic airways obstruction. A clinicopathological study of patients in London and Chicago. Lancet 1966;287:830–35.
    1. Oelsner EC, Hoffman EA, Folsom AR, et al. Association between emphysema-like lung on cardiac computed tomography and mortality in persons without airflow obstruction: a cohort study. Ann Intern Med 2014;161(12):863–73.
    1. Woodruff PG, Barr RG, Bleecker E, et al. Clinical Significance of Symptoms in Smokers with Preserved Pulmonary Function. N Engl J Med 2016;374(19):1811–21.
    1. Balte PP, Chaves PHM, Couper DJ, et al. Association of Nonobstructive Chronic Bronchitis With Respiratory Health Outcomes in Adults. JAMA Intern Med 2020;180(5):676–86.
    1. McAllister D, Ahmed FS, Austin JHM, et al. Emphysema predicts hospitalisation and incident airflow obstruction among older smokers: a prospective cohort study. PLoS ONE 2014;9(4):e93221.
    1. Oelsner EC, Carr JJ, Enright PL, et al. Per cent emphysema is associated with respiratory and lung cancer mortality in the general population: a cohort study. Thorax 2016;71(7):624–32.
    1. Ash SY, San José Estépar R, Fain SB, et al. Relationship between Emphysema Progression at CT and Mortality in Ever-Smokers: Results from the COPDGene and ECLIPSE Cohorts. Radiology 2021;299(1):222–31.
    1. Leopold JG, Gough J. The Centrilobular Form of Hypertrophic Emphysema and its Relation to Chronic Bronchitis. Thorax 1957;12(3):219–35.
    1. Edge J, Simon G, Reid L. Peri-acinar (paraseptal) emphysema: its clinical, radiological, and physiological features. British Journal of Diseases of the Chest 1966;60(1):10–18.
    1. Barr RG, Berkowitz EA, Bigazzi F, et al. A combined pulmonary-radiology workshop for visual evaluation of COPD: study design, chest CT findings and concordance with quantitative evaluation. COPD 2012;9(2):151–9.
    1. Smith BM, Austin JHM, Newell JD, Jr., et al. Pulmonary emphysema subtypes on computed tomography. The MESA COPD Study. Am J Med 2014;127:94.e7–23.
    1. Lynch DA, Austin JH, Hogg JC, et al. CT-Definable Subtypes of Chronic Obstructive Pulmonary Disease: A Statement of the Fleischner Society. Radiology 2015;277(1):192–205.
    1. Hinton GS, Terrence (1999). Unsupervised Learning: Foundations of Neural Computation. MIT Press. 1999.
    1. Castaldi PJ, Boueiz A, Y J, et al. Machine learning characterization of COPD subtypes: insights from the COPDGene study. Chest 2020;57(5):1147–57.
    1. Delgado-Eckert E, James A, Meier-Girard D, et al. Lung function fluctuation patterns unveil asthma and COPD phenotypes unrelated to type 2 inflammation. Journal of Allergy and Clinical Immunology 2021;48(2):407–19.
    1. Augustin IM, Spruit MA, Houben-Wilke S, et al. The respiratory physiome: clustering based on a comprehensive lung function assessment in patients with COPD. PLoS One 2018;13(9):e0201593.
    1. Gillenwater LA, Helmi S, Stene E, et al. Multi-omics subtyping pipeline for chronic obstructive pulmonary disease. PloS one 2021;16(8):e0255337.
    1. Zou C, Li F, Choi J, et al. Longitudinal imaging-based clusters in former smokers of the copd cohort associate with clinical characteristics: The subpopulations and intermediate outcome measures in copd study (SPIROMICS). International Journal of Chronic Obstructive Pulmonary Disease 2021;16:1477–96.
    1. Young AL, Bragman FJ, Rangelov B, et al. Disease progression modeling in chronic obstructive pulmonary disease. American Journal of Respiratory and Critical Care Medicine 2020;201(3):294–302.
    1. Yang J, Angelini ED, Balte PP, et al. Novel Subtypes of Pulmonary Emphysema Based on Spatially-Informed Lung Texture Learning: the Multi-Ethnic Study of Atherosclerosis (MESA) COPD Study. IEEE Transactions on Medical Imaging 2021;:1–1. doi: 10.1109/TMI.2021.3094660
    1. Couper D, LaVange LM, Han M, et al. Design of the subpopulations and intermediate outcomes in COPD study (SPIROMICS). Thorax 2014;69(5):492–95.
    1. Aaron CP, Hoffman EA, Kawut SM, et al. Ambient air pollution and pulmonary vascular volume on computed tomography: the MESA Air Pollution and Lung cohort studies. Eur Respir J 2019;53(6)
    1. Kaufman JD, Adar SD, Allen RW, et al. Prospective study of particulate air pollution exposures, subclinical atherosclerosis, and clinical cardiovascular disease: The Multi-Ethnic Study of Atherosclerosis and Air Pollution (MESA Air). Am J Epidemiol 2012;176(9):825–37.
    1. Regan EA, Hokanson JE, Murphy JR, et al. Genetic epidemiology of COPD (COPDGene) study design. COPD: Journal of Chronic Obstructive Pulmonary Disease 2011;7(1):32–43.
    1. Sieren JP, Newell JD Jr, Barr RG, et al. SPIROMICS protocol for multicenter quantitative computed tomography to phenotype the lungs. American Journal of Respiratory and Critical Care Medicine 2016;194(7):794–806.
    1. Hoffman EA, Jiang R, Baumhauer H, et al. Reproducibility and Validity of Lung Density Measures from Cardiac CT Scans—The Multi-Ethnic Study of Atherosclerosis (MESA) Lung Study. Journal of Academic Radiology 2009;16:689–99.
    1. Gevenois PA, De Maertelaer V, De Vuyst P, et al. Comparison of computed density and macroscopic morphometry in pulmonary emphysema. American Journal of Respiratory and Critical Care Medicine 1995;152(2):653–57.
    1. Hoffman EA, Ahmed FS, Baumhauer H, et al. Variation in the percent of emphysema-like lung in a healthy, nonsmoking multiethnic sample. The MESA lung study. Annals of the American Thoracic Society 2014;11(6):898–907.
    1. Gangeh MJ, Sorensen L, Shaker SB, et al. A texton-based approach for the classification of lung parenchyma in CT images. International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI), 2010:595–602.
    1. Gorelick L, Galun M, Sharon E, et al. Shape representation and classification using the Poisson equation. IEEE Transactions on Pattern Analysis and Machine Intelligence 2006;28:1991–2005.
    1. Rosvall M, Bergstrom CT. Maps of random walks on complex networks reveal community structure. Proceedings of the National Academy of Sciences 2008;105:1118–23.
    1. Yang J, Angelini ED, Smith BM, et al. Explaining Radiological Emphysema Subtypes with Unsupervised Texture Prototypes: MESA COPD Study. Medical computer vision and Bayesian and graphical models for biomedical imaging : MICCAI 2016 international workshop, MCV and BAMBI, Athens, Greece, October 21, 2016 : revised selected papers 2017;2017:69–80. doi: 10.1007/978-3-319-61188-4_7
    1. Häme Y, Angelini EA, Parikh ME, et al. Sparse sampling and unsupervised learning of lung texture patterns in pulmonary emphysema: MESA COPD study. IEEE International Symposium on Biomedical Imaging 2015:109–13.
    1. Yang J, Angelini ED, Balte PP, et al. Unsupervised Discovery of Spatially-Informed Lung Texture Patterns for Pulmonary Emphysema: The MESA COPD Study. Medical image computing and computer-assisted intervention : MICCAI International Conference on Medical Image Computing and Computer-Assisted Intervention 2017;10433:116–24. doi: 10.1007/978-3-319-66182-7_14 [published Online First: 2018/01/23]
    1. Maaten LV, Hinton G. Visualizing data using t-SNE. Journal of Machine Learning Research 2008;9:2579–605.
    1. Ganin Y, Lempitsky V. Unsupervised domain adaptation by backpropagation. International Conference on Machine Learning, 2015:1180–89.
    1. Sack CS, Doney BC, Podolanczuk AJ, et al. Occupational Exposures and Subclinical Interstitial Lung Disease. The MESA (Multi-Ethnic Study of Atherosclerosis) Air and Lung Studies. Am J Respir Crit Care Med 2017;196(8):1031–39.
    1. Kim V, Davey A, Comellas AP, et al. Clinical and computed tomographic predictors of chronic bronchitis in COPD: a cross sectional analysis of the COPDGene study. Respiratory Research 2014;15(1):52.
    1. Miller MR, Crapo R, Hankinson J, et al. General considerations for lung function testing. European Respiratory Journal 2005;26:153–61.
    1. McDonough JE, Yuan R, Suzuki M, et al. Small-airway obstruction and emphysema in chronic obstructive pulmonary disease. N Engl J Med 2011;365(17):1567–75.
    1. Galban CJ, Han MK, Boes JL, et al. Computed tomography-based biomarker provides unique signature for diagnosis of COPD phenotypes and disease progression. Nature medicine 2012;18(11):1711–5.
    1. Smith B, Kirby M, Hoffman E, et al. MESA Lung, CanCOLD, and SPIROMICS investigators. Association of dysanapsis with chronic obstructive pulmonary disease among older adults. JAMA 2020;323(22):2268–80.
    1. Oelsner EC, Loehr LR, Henderson AG, et al. Classifying Chronic Lower Respiratory Disease Events in Epidemiologic Cohort Studies. Annals of the American Thoracic Society 2016;13(7):1057–66.
    1. Consortium GTEx. The GTEx Consortium atlas of genetic regulatory effects across human tissues. Science 2020;369(6509):1318–30.
    1. Giambartolomei C, Vukcevic D, Schadt E, et al. Bayesian test for colocalisation between pairs of genetic association studies using summary statistics. PLoS Genet 2014;10(5):e1004383.
    1. Wallace C Eliciting priors and relaxing the single causal variant assumption in colocalisation analyses. PLoS Genet 2020;16(4):e1008720.
    1. Herman AI, DeVito EE, Jensen KP, et al. Pharmacogenetics of nicotine addiction: role of dopamine. Pharmacogenomics 2014;15(2):221–34.
    1. Matsuyama N, Shibata S, Matoba A, et al. The dopamine D 1 receptor is expressed and induces CREB phosphorylation and MUC5AC expression in human airway epithelium. Respiratory Research 2018;19(1):53.
    1. Kesimer M, Applin M, Ford A, et al. Airway Mucin Concentration as a Marker of Chronic Bronchitis. New England Journal of Medicine 2017;377:911–22.
    1. Kesimer M, Smith BM, Ceppe A, et al. Mucin Concentrations and Peripheral Airways Obstruction in COPD. Am J Respir Crit Care Med 2018. doi: 10.1164/rccm.201806-1016LE [published Online First: 2018/08/22]
    1. Martinez FJ, Han MK, Allinson JP, et al. At the Root: Defining and Halting Progression of Early Chronic Obstructive Pulmonary Disease. Am J Respir Crit Care Med 2018;197(12):1540–51.
    1. Burrows B, Kettel LJ, Niden AH, et al. Patterns of cardiovascular dysfunction in chronic obstructive pulmonary disease. New Engl J Med 1972;17:912–18.
    1. Hueper K, Vogel-Claussen J, Parikh MA, et al. Pulmonary Microvascular Blood Flow in Mild Chronic Obstructive Pulmonary Disease and Emphysema. The MESA COPD Study. Am J Respir Crit Care Med 2015;192(5):570–80.
    1. Thomashow MA, Shimbo D, Parikh MA, et al. Endothelial microparticles in mild COPD and emphysema: The MESA COPD Study. Am J Respir Crit Care Med 2013;188:60–8.
    1. Wang M, Aaron CP, Madrigano J, et al. Association Between Long-term Exposure to Ambient Air Pollution and Change in Quantitatively Assessed Emphysema and Lung Function. JAMA 2019;322(6):546–56.
    1. Auerback O, Hammond EC, Garfinkel L. Relationship of smoking and age to emphysema. Whole-lung section study. New England Journal of Medicine 1972;286(16):853–57.
    1. Bickerman HA. Senile emphysema. J Am Geriatr Soc 1956;4(6):526–34.
    1. Schiffers C, Lundblad LKA, Hristova M, et al. Downregulation of DUOX1 function contributes to aging-related impairment of innate airway injury responses and accelerated senile emphysema. Am J Physiol Lung Cell Mol Physiol 2021;321(1):L144–l58.
    1. Wicher SA, Roos BB, Teske JJ, et al. Aging increases senescence, calcium signaling, and extracellular matrix deposition in human airway smooth muscle. PLoS One 2021;16(7):e0254710.
    1. Cheng CY, Yamashiro K, Chen LJ, et al. New loci and coding variants confer risk for age-related macular degeneration in East Asians. Nature Communications 2015;6:6063.
    1. Fishilevich S, Zimmerman S, Kohn A, et al. Genic insights from integrated human proteomics in GeneCards. Database 2016;2016
    1. Li ZG, Scott MJ, Brzoska T, et al. Lung epithelial cell-derived IL-25 negatively regulates LPS-induced exosome release from macrophages. Military Medical Research 2018;5(1):24.
    1. Srivastava A, Amreddy N, Razaq M, et al. Exosomes as Theranostics for Lung Cancer. Advances in cancer research 2018;139:1–33.
    1. Lin H, Jiang S. Combined pulmonary fibrosis and emphysema (CPFE): an entity different from emphysema or pulmonary fibrosis alone. Journal of Thoracic Disease 2015;7(4):767.
    1. Wan ES, Castaldi PJ, Cho MH, et al. Epidemiology, genetics, and subtyping of preserved ratio impaired spirometry (PRISm) in COPDGene. Respir Res 2014;15:89.
    1. Ladizinski B, Sankey C. Vanishing lung syndrome. New England Journal of Medicine 2014;370(9):e14.
    1. Li F, Choi J, Zou C, et al. Latent traits of lung tissue patterns in former smokers derived by dual channel deep learning in computed tomography images. Scientific Reports 2021;11(1):4916.
    1. Haghighi B, Choi S, Choi J, et al. Imaging-based clusters in former smokers of the COPD cohort associate with clinical characteristics: the SubPopulations and intermediate outcome measures in COPD study (SPIROMICS). Respiratory Research 2019;20(1):153.
    1. Binder P, Batmanghelich NK, Estépar RSJ, et al. Unsupervised discovery of emphysema subtypes in a large clinical cohort. International Workshop on Machine Learning in Medical Imaging: Springer, 2016:180–87.

Source: PubMed

Sponsorer och medarbetare

Medicinska tillstånd

Läkemedelsinterventioner

3
Prenumerera