CT of Viral Lower Respiratory Tract Infections in Adults: Comparison Among Viral Organisms and Between Viral and Bacterial Infections

Cardiopulmonary Imaging Original Research Miller et al. CT of Viral Lower Respiratory Tract Infections Cardiopulmonary Imaging Original Research Wallace T. Miller Jr 1 Timothy J. Mickus 2 Eduardo Barbosa Jr 1 Christopher Mullin 3 Vivanna M. Van Deerlin 4 Kevin T. Shiley 5 Miller Jr WT, Mickus TJ, Barbosa Jr E, Mullin C, Van Deerlin VM, Shiley KT Keywords: adenovirus, CT, influenza virus, respiratory syncytial virus, viral pneumonia DOI:10.2214/AJR.11.6501 Received January 15, 2011; accepted after revision April 7, 2011. 1 Department of Radiology, University of Pennsylvania School of Medicine, Silverstein 1, 3400 Spruce St, Philadelphia, PA 19104. Address correspondence to W. T. Miller, Jr. (millerw@uphs.upenn.edu). 2 Department of Radiology, Allegheny General Hospital, Pittsburgh, PA. 3 Department of Medicine, Columbia University Medical Center, New York, NY. 4 Department of Pathology and Laboratory Medicine, University of Pennsylvania School of Medicine, Philadelphia, PA. 5 Division of Infectious Diseases, Mercy Hospital of Buffalo, Buffalo, NY. AJR 2011; 197:1088 1095 0361 803X/11/1975 1088 American Roentgen Ray Society CT of Viral Lower Respiratory Tract Infections in Adults: Comparison Among Viral Organisms and Between Viral and Bacterial Infections OBJECTIVE. We retrospectively compared the CT findings of consecutive viral and bacterial lower respiratory tract infections (LRTIs) to determine their imaging appearance and any definable differences among the causative viruses and between the viral and bacterial infections. MATERIALS AND METHODS. Imaging features of LRTI caused by influenza virus, respiratory syncytial virus (RSV), parainfluenza, adenovirus, and bacteria over a 33-month period were reviewed by three radiologists blinded to clinical and diagnostic information. Individual CT features and the dominant pattern of infection were recorded for each examination. Imaging characteristics were compared among the four respiratory viruses and between viral and bacterial infections. RESULTS. One hundred fifteen chest CT scans were analyzed (60 influenza virus, 19 RSV, 10 adenovirus, four parainfluenza virus, and 22 bacterial pneumonia LRTIs). Individual imaging findings and imaging patterns were seen in similar frequencies when we compared viral and bacterial LRTIs, with the exception of the diffuse airspace pattern, which was seen more frequently in bacterial infections. Although there was overlap in the imaging appearance of individual viruses, RSV and adenovirus tended to have characteristic imaging appearances. RSV presented with an airway-centric pattern of disease (13/19 cases [68%]) characterized by varying mixtures of tree-in-bud opacities and bronchial wall thickening, with or without peribronchiolar consolidation. Adenovirus typically appeared as multifocal consolidation or ground-glass opacity without airway inflammatory findings (7/10 cases [70%]). CONCLUSION. There is considerable overlap in the imaging appearance of viral and bacterial respiratory infections. However, some characteristic differences can be seen, especially with RSV and adenovirus infections. C T of common community-acquired viral lower respiratory tract infections (LRTIs) is still in its infancy, with only a few previously published studies comprising approximately 250 cases [1 19]. However, to our knowledge, direct comparisons of the imaging features of the different viral causes of LRTI have not been performed. Furthermore, the imaging features of viral LRTI have not been compared with bacterial infections, to our knowledge. We performed an analysis of CT images in patients with a proved diagnosis of infection with the four most common viruses responsible for community-acquired viral LRTI: influenza virus, respiratory syncytial virus (RSV), parainfluenza virus, and adenovirus. We had three objectives: first, to determine whether CT patterns reported in prior smaller series, including our own, would hold true in a larger cohort of patients and with multiple reviewers; second, to determine whether individual viruses showed distinctive imaging appearances; and third, to determine whether viral infections could be distinguished from bacterial infections. Materials and Methods Study Design The study was approved by the University of Pennsylvania Medical Center s institutional review board and was HIPAA compliant. Informed consent was waived because of the purely retrospective nature of the project. We defined a patient with LRTI as one with lower respiratory tract symptoms (i.e., new onset cough, dyspnea, or sputum production) and laboratory evidence of viral or bacterial respiratory infection. For inclusion into the study, 1088 AJR:197, November 2011

CT of Viral Lower Respiratory Tract Infections TABLE 1: CT Protocols and Scanners Used in 115 CT Scans Protocol, Scanner No. of Scans Protocol Specifics Unenhanced helical CT Sensation 64, Siemens Healthcare 15 Definition 64, Siemens Healthcare 8 Sensation 16, Siemens Healthcare 7 Volume Zoom, Siemens Healthcare 1 LightSpeed QXi, GE Healthcare 8 Enhanced helical CT Sensation 64, Siemens Healthcare 5 LightSpeed QXi, GE Healthcare 2 PE protocol 1 Sensation 64, Siemens Healthcare 6 Definition 64, Siemens Healthcare 1 Sensation 16, Siemens Healthcare 7 Volume Zoom, Siemens Healthcare 4 PE protocol 2 Sensation 64, Siemens Healthcare 11 Definition 64, Siemens Healthcare 5 Sensation 16, Siemens Healthcare 12 Axial high-resolution CT Sensation 4, Siemens Healthcare 1 LightSpeed QXi, GE Healthcare 6 Spiral high-resolution CT Sensation 64, Siemens Healthcare 9 Definition 64, Siemens Healthcare 7 Note PE = pulmonary embolism. subjects had to meet three criteria: first, be an adult with clinical symptoms of an acute LRTI presenting as an outpatient between November 1, 2005, and July 31, 2008; second, have a polymerase chain reaction (PCR) of respiratory secretions positive for influenza virus, RSV, adenovirus, or parainfluenza virus or have respiratory cultures positive for a known bacterial pathogen; and third, have undergone a chest CT within 72 hours of the PCR. The first two criteria are the accepted clinical criteria for the diagnosis of viral LRTIs [20 22]. It should be noted that this definition is inclusive of viral bronchitis, viral bronchiolitis, and viral pneumonia. We have relied on this definition of viral LRTI because it excludes imaging features as criteria for disease, because it was our goal to identify imaging features of these infections. The study was performed at two affiliated university-based hospitals in the northeastern United States: a 725-bed academic tertiary care center and an urban 324-bed community hospital. Subjects were initially identified through a prospectively acquired database of patients who underwent testing for influenza virus, RSV, adenovirus, and parainfluenza virus infection. The PCR was performed as a partially multiplexed real-time reverse-transcription PCR for RNA viruses or as real-time PCR for DNA viruses on samples obtained from nasopharyngeal swabs or bronchoalveolar lavage fluid. Assays were performed on an ABI 7900 with TaqMan probes (both from Applied Biosystems). Beta-2 microglobulin was used as a control for the presence of inhibitors or for poor sample quality. The radiology information system was retrospectively reviewed to determine whether a chest CT scan had occurred within 72 hours of assay collection. The medical records of subjects with CT images were reviewed in detail to identify those with symptoms of LRTI. Patients meeting these three criteria were subdivided into four groups: group 1, those with evidence of viral infection without concurrent bacterial infection; group 2, those with evidence of bacterial infection without concurrent viral infection; group 3, those with evidence of concurrent viral and bacterial infection; and group 4, those without evidence of either bacterial or viral infection. Groups 3 and 4 were excluded from further analysis. Bacterial LRTI was defined as the presence of a positive respiratory culture with a bacterial lung pathogen, by blood cultures growing Streptococcus pneumoniae, or by a positive Legionella species or pneumococcal urinary antigen result. Cultures with normal oral flora were not considered to be evidence of infection. Patients with preexisting diffuse lung disease were excluded from analysis because it was not possible to reliably distinguish findings related to viral LRTI from those caused by the other disease. Forty-two of the CT scans reviewed were included in an earlier study of viral respiratory tract infections [7]. CT Analysis CT protocols used during the study and the relative frequency of protocols and scanners are detailed in Table 1. IV contrast agent was administered for 59 of 115 examinations. Standard heli- One-mm contiguous spiral, deep inspiration, no contrast agent administered, viewed as 5-mm contiguous slices One-mm contiguous spiral, deep inspiration, 100 ml of iohexol (300 mg I/mL [Omnipaque 300, GE Healthcare]) administered at 1 2 ml/s, viewed as 5-mm contiguous slices One-mm contiguous spiral, deep inspiration, no contrast agent administered, viewed as 5-mm contiguous slices; or 1-mm contiguous spiral, deep inspiration, 100 ml iohexol (350 mg I/mL [Omnipaque 350, Omnipaque, GE Healthcare]) administered at 3 5 ml/s, viewed at 1 3-mm contiguous slices One-mm contiguous spiral, deep inspiration, 100 ml iohexol (350 mg I/mL) administered at 3 5 ml/s, viewed at 1 3-mm contiguous slices Five-mm spiral, contiguous; 1-mm axial, 10-mm intervals deep inspiration; 1-mm axial, 10-mm intervals end expiration; or stationary 1-mm axial, imaged every second during forced expiration at three locations: aortic arch, carina, and 1 cm above diaphragm One-mm contiguous spiral, deep inspiration, viewed as 5-mm contiguous slices and 1-mm slice at 10-mm intervals and 5-mm coronal reconstructions; 1-mm contiguous spiral, end expiration, viewed as 5-mm contiguous slices and 1-mm slice at 10-mm intervals and 5-mm coronal reconstructions AJR:197, November 2011 1089

Miller et al. TABLE 2: Demographic Features of 115 Patients With Lower Respiratory Tract Infection Characteristic Viral Infection (n = 93) Bacterial Infection (n = 22) Age (y), mean (range) 55 (18 85) 57 (31 82) Female sex 51 (55) 8 (36) Immunocompromised 42 (45) 12 (54) Cancer 22 (24) 4 (18) Leukemia 9 (10) 1 (5) Lymphoma 4 (4) 0 Multiple myeloma 5 (5) 1 (5) Other 4 (4) 2 (9) Transplantation 12 (13) 7 (32) HIV 4 (4) 0 Steroids 3 (3) 1 (5) Other 1 (1) 0 (0) Note Data are no. (%) of subjects, except where noted otherwise. cal CT examinations used 100 ml of iohexol 300 mg I/mL (Omnipaque 300, GE Healthcare), and pulmonary embolism (PE) protocol examinations used 100 ml of iohexol 350 mg I/mL (Omnipaque 350, GE Healthcare). Examinations were retrospectively and independently reviewed by three radiologists with 19 years (reader 1), 2 years (reader 3), and 1 year (reader 2) of experience in thoracic imaging. All reviewers were blinded to clinical information and the original CT reports. CT images were analyzed in a randomized manner at least 6 months after the examination date to eliminate recall bias. Examinations were evaluated for the presence or absence of tree-in-bud opacities, bronchial wall thickening, bronchiectasis, ground-glass opacities, and airspace consolidation. There are no well-established criteria for bronchial wall thickening. We initially attempted to measure the wall thickness of the fourth- or fifth-generation bronchus. However, these bronchi walls were typically only 1 2 pixels in thickness (at average in-plane spatial resolution of 0.3 mm); therefore, objective measures were considered too inaccurate. Thus, the determination of bronchial wall thickening was made when the reviewing radiologist initially suspected the presence of bronchiectasis but then on comparing the caliber of the bronchus to that of the adjacent artery determined that there was no bronchial dilatation. Thus, the bronchus seemed more prominent than normal but did not show features of bronchial dilation. The presence or absence of pleural effusions was also noted. After evaluating specific imaging findings, the interpreting radiologist made a determination of the dominant imaging pattern, which represented the most extensive and prominent features of each CT scan. On the basis of our earlier investigations, these dominant patterns were categorized into seven groups: pattern 1, bronchitis, which was characterized by the presence of bronchial wall thickening; pattern 2, bronchiolitis, which was characterized by the presence of tree-in-bud opacities with or without bronchial wall thickening; pattern 3, bronchopneumonia, which was characterized by the presence of tree-in-bud opacities or bronchial wall thickening, or both, with bronchocentric areas of ground-glass opacity or consolidation, or both; pattern 4, multifocal pneumonia, which was characterized by multiple regions of consolidation or ground-glass opacity, or both, with intervening regions of normal lung parenchyma and without evidence of bronchiolitis; pattern 5, unifocal infection, which was characterized by unifocal ground-glass opacity, tree-in-bud opacities, or consolidation; pattern 6, diffuse airspace disease, which was characterized by relatively uniform ground-glass opacity or consolidation, or both, throughout both lungs; and pattern 7, normal examination or abnormalities unrelated to infection (e.g., emphysema or small linear scars) [7]. Patterns 1 3 were further grouped together as airwaycentric disease. Statistical Analysis Fisher exact test was used to compare categoric data. Reader agreement was measured using the kappa statistic. All p values reported are two sided, with a p value less than 0.05 considered statistically significant. Results A total of 2518 patients were tested for respiratory viral infection, and 952 (37.8%) had assays positive for respiratory virus infection; of those 952 patients, 123 (12.9%) had a chest CT performed within 72 hours of the viral assay. Twelve patients had severe preexisting extensive parenchymal lung disease and were excluded from further analysis. Eighteen subjects had evidence of concurrent viral and bacterial infection and were excluded from further analysis. Ninety-three of the virus-positive subjects had evidence of viral infection without bacterial infection and made up our viral infection group. These consisted of 60 patients (65%) with influenza virus infection, 19 patients (21%) with RSV infection, 10 patients (11%) with adenovirus infection, and four patients (4%) with parainfluenza virus infection. Twenty-two of the patients who tested negative for respiratory virus infection met our criteria for community-acquired bacterial infection and made up our bacterial-only LRTI group. Organisms included Pseudomonas aeruginosa (n = 5), Staphylococcus aureus (n = 11), Klebsiella pneumoniae (n = 3), Streptococcus pneumoniae (n = 2), and Achromobacter xylosoxidans (n = 1). Demographic features of the viral and bacterial groups are listed in Table 2. The demographic characteristics of the two groups were similar (all p > 0.05). There was no significant difference in the presenting symptoms, including cough, dyspnea, chest pain, and wheezing, for all causative organisms (all p > 0.05). Reader Agreement There was strong agreement among the three readers about the presence or absence of imaging findings, with the exception of bronchial wall thickening. Excluding bronchial wall thickening, all three readers agreed on the absence or presence of a finding (universal agreement) for 80 90% of findings (Table 3). There was universal agreement on the presence or absence of bronchial wall thickening in only 80 of 133 cases (60%). Kappa values in most cases showed moderate (κ = 0.41 0.60), substantial (κ = 0.61 0.80), or almost perfect (κ = 0.81 1.0) agreement among reviewers. Exceptions to this pattern included only fair agreement concerning the presence of bronchial wall thickening between reader 3 and readers 1 and 2, only fair agreement concerning the presence of bronchiectasis between reader 2 and readers 1 and 3, and only slight agreement between readers 1 and 3 concerning the presence of focal pneumonia. When imaging patterns rather than individual findings were compared, reader agree- 1090 AJR:197, November 2011

CT of Viral Lower Respiratory Tract Infections TABLE 3: Reader Agreement for Imaging Findings and Patterns of 133 CT Examinations Finding ment was very high, with an overall kappa value of 0.79. Kappa values of individual patterns between readers are listed in Table 3. All three reviewers agreed on the imaging pattern in 102 of 133 examinations (77%). Imaging Findings for Viruses as a Group Because of the exceptional reader agreement, all results based on individual readers were the same as the aggregated results. For simplicity, we report the mean findings for the three reviewers. Individual imaging findings are summarized in Table 4. A normal CT scan or one with findings unrelated to viral infection (e.g., emphysema or nonspecific scarring) was observed in 33 of 93 patients (36%). Evidence of airway inflammation was found in approximately one third of cases, manifested by bronchial wall thickening (29/93 [32%]) or tree-in-bud opacities (29/93 [32%]) (Figs. 1 4). Airspace filling was apparent in approximately one fourth of patients, primarily manifesting as multifocal ground-glass opacities (20/93 [22%]) or multifocal consolidation (25/93 [26%]) or a combination of both (Figs. 5 and 6). Diffuse TABLE 4: Findings and Dominant CT Pattern in 115 Cases of Community-Acquired Respiratory Tract Infection Finding Finding and Pattern RSV (n = 19) Parainfluenza Virus (n = 4) Adenovirus (n = 10) Influenza Virus (n = 60) All Viruses (n = 93) No findings 4 (21) 1 (25) 2 (20) 26 a (43) 33 (35) 2 (9) Bronchiectasis 1 (5) 0 (0) 1 (10) 5 (8) 7 (8) 3 (14) Tree-in-bud 10 b (53) 2 (50) 1 (10) 16 (27) 29 (31) 3 (14) Bronchial wall thickening 8 c (42) 2 (50) 1 (10) 19 (32) 29 (31) 6 (27) Ground-glass opacity 3 (16) 1 (25) 6 d (60) 10 (17) 20 (22) 10 (45) Multifocal consolidation 4 (21) 1 (25) 7 d (70) 14 (23) 25 (27) 8 (36) Focal consolidation 1 (5) 0 (0) 0 (0) 5 (8) 6 (6) 2 (9) Pleural effusion 7 (37) 1 (25) 3 (30) 9 (15) 20 (22) 9 (41) Pattern Finding or Pattern Universal Agreement a Reader 1 and Reader 2 κ Reader 1 and Reader 3 Reader 2 and Reader 3 Bronchiectasis 112 (84) 0.34 0.43 0.65 Tree-in-bud 107 (80) 0.62 0.53 0.76 Bronchial wall thickening 80 (60) 0.59 0.21 0.34 Ground-glass opacity 108 (81) 0.69 0.43 0.69 Multifocal consolidation 109 (82) 0.72 0.84 0.73 Focal consolidation 117 (88) 0.42 0.15 0.42 Pleural effusion 120 (90) 0.80 0.80 0.94 Pattern Airway 0.80 0.71 0.85 Multifocal airspace 0.87 0.78 0.88 Focal airspace 0.79 0.58 0.64 Diffuse airspace 0.66 0.66 1.0 Normal or other b 0.86 0.81 0.76 a Data are no. (%) of cases where all three observers agreed on the presence or absence of a finding. b Normal study or other findings unrelated to viral infection, such as emphysema or focal scarring. Airway 13 b (68) 2 (50) 0 16 (27) 31 (33) 4 (18) Multifocal airspace 2 (11) 1 (25) 7 d (70) 12 (20) 22 (24) 5 (23) Diffuse airspace 0 0 1 (10) 1 (2) 2 (2) 6 e (27) Focal airspace 0 0 0 5 (8) 5 (5) 1 (5) Other/normal 4 (21) 1 (25) 2 (20) 26 a (43) 33 (35) 6 (27) Bacteria (n = 22) Note All numbers are averages for the three reviewers. RSV = respiratory syncytial virus. a Influenza virus infection presented with normal CT findings at significantly higher frequencies compared with all other viruses for reviewers 1 and 3 (p = 0.02 0.03). b RSV infection presented with tree-in-bud significantly more often than all other viruses for all three reviewers (p = 0.004 to < 0.001) and presented with evidence of airway-centric disease significantly more often than all other viruses for all three reviewers (p = 0.004 to < 0.001). c RSV infection presented with bronchial wall thickening significantly more often than all other viruses for reviewer 1 (p = 0.05). d Adenovirus infection had a greater frequency of multifocal airspace consolidation (p = 0.004 0.02) and diffuse or multifocal ground-glass opacity (p = 0.004 0.01) compared with all other viruses for all three reviewers, and adenovirus infection was more likely to appear as multifocal pneumonia compared with all other viruses for all three reviewers (p = 0.001). e Bacterial infections appeared as diffuse airspace disease significantly more frequently than individual viruses and all viruses as a group for all three reviewers (p = 0.002 0.009). AJR:197, November 2011 1091

Miller et al. Fig. 1 41-year-old HIV-positive man with 1 month of cough, dyspnea, and night sweats. CT image shows multiple tree-in-bud opacities in lungs bilaterally, associated with mild bronchial wall thickening (arrow). Diagnosis was bronchiolitis as a result of respiratory syncytial virus infection. Fig. 4 51-year-old man with fever, cough, and dyspnea. Diagnosis was bronchopneumonia as a result of respiratory syncytial virus infection. CT image shows multiple small tree-in-bud opacities and bronchial wall thickening indicating bronchiolitis. However, there are also regions of confluent groundglass opacities and patchy consolidation indicating extension of bronchiolitis into bronchopneumonia. ground-glass opacity or consolidation across the entire lungs was uncommon (2/93 [2%]). Rarely, viral pneumonia appeared as unifocal regions of ground-glass opacity or consolidation (6/92 [7%]). When viewed as a group, viral LRTIs tended to cause four imaging patterns: airway-centric disease (bronchitis, bronchiolitis, or bronchopneumonia) (Figs. 1 4); multifocal pneumonia (Figs. 5 and 6); unifocal infection manifested as a single focal region of consolidation, groundglass opacity, or tree-in-bud opacities (Table 4); or a normal study with no imaging findings related to infection. Overall, 33% (31/93) of viral LRTIs had an airway-centric pattern of disease, 24% (22/93) had an appearance of multifocal pneumonia, 5% (5/93) had an appearance of a unifocal infection, and 36% (33/93) had no imaging evidence of lower respiratory infection. Rarely, viral infection appeared as diffuse ground-glass opacity or consolidation, nearly Fig. 2 49-year-old woman with cough and fever. CT image shows predominantly bronchial wall thickening and few tree-in-bud opacities in lungs bilaterally, indicating predominantly bronchitis with mild parainfluenza bronchiolitis. Note similarity to respiratory syncytial virus bronchiolitis in Figure 1. Fig. 5 50-year-old woman with chronic obstructive pulmonary disease who presented with cough, dyspnea, fever, and respiratory failure. Diagnosis was multifocal pneumonia due to influenza A virus infection. CT image shows multiple areas of consolidation and ground-glass opacity without airway abnormalities as a result of influenza A pneumonia. uniformly distributed across the lungs bilaterally (2/93 [2%]). Viral Infection Compared With Bacterial Infection Viral LRTIs as a group had an imaging appearance similar to that of bacterial infections in our population. The only imaging feature that was significantly different between the two types of infection was the frequency of diffuse airspace disease. Diffuse airspace opacification with ground-glass opacity or consolidation, or both, was seen with a much higher frequency in bacterial infection (6/22 [27%]) than in viral infections (2/93 [2%]), a difference that was statistically significant for all three readers (p = 0.002 0.009). Viral infections were also more likely to present with a normal CT examination or one without imaging finding of infection (i.e., emphysema) than were bacterial infections (p = 0.003 0.01 for all three reviewers). Fig. 3 49-year-old man with influenza A bronchiolitis and leukemia after bone marrow transplant with fever and cough. CT image shows multiple tree-in-bud opacities in lungs bilaterally, associated with mild bronchial wall thickening similar to respiratory syncytial virus and parainfluenza bronchiolitis in Figures 1 and 2. Fig. 6 42-year-old woman with fever, cough, and dyspnea. Diagnosis was multifocal pneumonia as a result of adenovirus infection. CT image shows multiple areas of consolidation and ground-glass opacity as a result of adenovirus infection. Note similarity to influenza pneumonia in Figure 5. Imaging Findings for Individual Viruses Individual viruses tended to have a higher frequency of certain imaging findings and patterns (Table 4). The airway-centric pattern of disease characterized by varying degrees of bronchial wall thickening and treein-bud opacities with or without multifocal consolidation was the predominant pattern in RSV infection (13/19 cases [68%]), where it occurred significantly more often compared with all other viruses for all three readers (p = 0.004 to < 0.001) (Figs. 1 and 4). This difference was also significant when comparing RSV with adenovirus (p = 0.00005 0.001) and RSV with influenza virus (p = 0.00004 0.01) infections for all three reviewers. The airway-centric pattern was also significantly more frequent for RSV compared with bacterial infection (p = 0.0001 0.01) for all three reviewers. The imaging finding of treein-bud opacities was highly associated with 1092 AJR:197, November 2011

CT of Viral Lower Respiratory Tract Infections RSV infection for all three reviewers (p = 0.03 0.05), and bronchial wall thickening was associated with RSV infection by reader 1 (p = 0.05). Adenovirus infection characteristically presented as a multifocal pneumonia pattern manifest as multifocal consolidation or groundglass opacity, or both, a finding that also reached statistical significance when compared with all other viruses for all three readers (p = 0.001) (Fig. 6). This difference was also significant when comparing adenovirus with RSV (p = 0.0005) and adenovirus with influenza virus (p = 0.004) infections for all three reviewers. In our analysis, we distinguished between multifocal and diffuse airspace disease. Using this distinction, adenovirus was statistically more likely to cause multifocal consolidation than bacterial infections (p = 0.003) for all three reviewers. However, if we grouped diffuse and multifocal airspace opacities together, the difference between adenovirus and bacterial infection was not significant (p = 0.07). The individual finding of multifocal consolidation (p = 0.004 0.02) was seen more frequently with adenovirus than with other viruses but was not seen more frequently than bacterial infections (p = 0.06 0.12). Ground-glass opacity was seen more frequently with adenovirus than with other viral infections (p = 0.004 0.01) and bacterial infection (p = 0.05). Influenza virus infection was most likely to present with a normal CT scan compared with the other viral infections for reader 1 (p = 0.02) and reader 3 (p = 0.03) but not for reader 2. Influenza virus infection had the widest range of imaging appearances, which was nearly equally divided between normal examinations, airway-centric disease, and multifocal pneumonia (Figs. 3 and 5). We had very limited data on parainfluenza virus given our small sample size, but found that two of four cases presented with an airway-centric pattern of infection, one appeared as a multifocal pneumonia, and one had no findings associated with the infection (Fig. 2). Discussion Although there have been a variety of previous reports on the imaging features of individual viral LRTIs, a holistic understanding of the imaging manifestations of viral infections has yet to be stated. The current larger series confirms our previously reported findings [7] that viral LRTIs tend to take one of four imaging appearances: airway-centric disease characterized by varying combinations of tree-in-bud opacities, bronchial wall thickening, and peribronchiolar consolidation (31/93 [33%]) (Figs. 1 4); multifocal pneumonia, characterized by multifocal consolidation or ground-glass opacity, or both, without airway findings (22/93 [24%]) (Figs. 5 and 6); unifocal infection manifested as a single focal region of consolidation, groundglass opacity, or tree-in-bud opacities (5/93 [5%]); or normal study with no imaging findings related to infection (33/93 [36%]). Our earlier study had included only one reviewer, whereas the current study has three reviewers who had exceptional reader agreement. The strong agreement among readers indicates that these patterns can be reliably recognized by multiple individuals. The findings of the present study suggest that the imaging appearance of viral and bacterial LRTIs have considerable overlap and that, in any individual case, viral infections cannot reliably be distinguished from bacterial infections. Individual findings of tree-in-bud opacities, bronchial wall thickening, ground-glass opacity, focal and multifocal airspace opacification, and pleural effusions were seen in similar frequencies in both viral and bacterial infection groups (Fig 4). Furthermore, airway centric, multifocal airspace, and focal airspace patterns of disease were seen in similar frequencies in both viral and bacterial groups. The only pattern with significantly different frequency between viral and bacterial infections was the diffuse airspace pattern, which was seen more frequently in bacterial infections (p = 0.002 0.009). However, we would like to point out that we had a relatively low proportion of bacterial infections. The study design, which required all individuals to have viral respiratory swab testing, biased the study toward viral infections. It is conceivable that the imaging and clinical features of bacterial infections in patients with a clinical presentation leading to viral testing could be different than the imaging and clinical features of bacterial LRTI in the general population. If this notion is true, it could alter the relative differences between viral and bacterial infections. Most radiologists are biased toward emphasizing bacterial pneumonia over bacterial airways infections (i.e., bronchitis and bronchiolitis). Bacterial pneumonias generally cause unifocal, multifocal, or diffuse consolidation. In our population, this constituted 12 of 22 cases (55%) of bacterial infection. However, four of 22 patients (18%) had imaging features of bronchiolitis or bronchitis with bacterial infection. Furthermore, six of 22 cases (27%) had no imaging findings of infection, suggesting that they had very mild bacterial airways infection, which did not result in radiographic changes. This finding suggests one of two conclusions: first, our population of patients with bacterial infection was not representative of bacterial LRTI in the general population, or, two, there may be a higher frequency of bacterial small-airways infection, bronchitis, or bronchiolitis than is generally recognized by radiologists. We believe that the latter conclusion is probably correct. Viral infections with RSV and adenovirus tended to have characteristic imaging presentations. RSV is known to predominantly cause the clinical syndrome of bronchiolitis [23]. Our study has confirmed that the imaging features of RSV infection correspond to those of bronchiolitis, appearing as an airway-centric pattern of disease in 13 of 19 cases (68%) that was characterized by varying combinations of tree-in-bud opacities, bronchial wall thickening, and peribronchiolar consolidation. Similar findings have been reported in the literature [11, 12, 16]. This airway-centric pattern of disease was highly associated with RSV infection when compared with all other viral infections, and when compared with adenovirus and influenza virus infection individually. This pattern was also statistically more common with RSV infection than with bacterial infections. Clinically, most adenovirus infections result in a mild upper respiratory tract illness. However, several clinical series have shown adenovirus to be a cause of severe pneumonia, resulting in hospitalization and the need for mechanical ventilation [24]. Our results confirm that the typical imaging appearance of adenovirus infection is widespread multifocal pneumonia. Seventy percent of cases (7/10) presented with a multifocal pneumonia pattern that was characterized by multifocal regions of ground-glass opacity or consolidation without airway abnormalities. This pattern was strongly associated with adenovirus infection for all three observers when compared with all other viruses (p = 0.001) and when compared with RSV (p = 0.0005) and influenza virus (p = 0.004) individually. Other researchers have also identified multifocal ground-glass opacity or consolidation as the typical appearance of adenoviral LRTI [9, 13, 14]. Influenza virus infection had the most variable imaging appearance of the community AJR:197, November 2011 1093

Miller et al. acquired viruses, presenting with a variety of patterns, including airway centric (16/60 [27%]), multifocal airspace (12/60 [20%]), and normal and noninfectious patterns (26/60 [43%]). It was the virus most likely to present with a normal CT or one without imaging features of infection (p = 0.02 0.03). To many, the frequency of normal CTs might be counterintuitive. However, we believe that this result matches what we would expect on the basis of the clinical presentation of influenza virus infection. Compared with the other respiratory viruses, influenza virus infection is more likely to present with severe systemic symptoms, such as headache, myalgias, and prostration. This is especially true in nonimmunocompromised hosts. As a consequence, individuals with influenza virus infection can appear severely ill, but with limited respiratory symptoms. We suspect that this may have resulted in a higher frequency of imaging examinations without clinical signs localizing the disease to the thorax. Therefore, although patients are clinically ill, they are less likely to have imaging evidence of LRTI. Although influenza virus had a low frequency of the multifocal pneumonia pattern (20%), it remained the most common viral cause of this pattern seen in our study, where it accounted for 12 cases, as compared with seven cases of adenovirus infection and two cases of RSV infection. Thus, although influenza virus infection has a lower incidence of the multifocal pneumonia pattern than adenovirus infection, it remains the most common cause of viral multifocal pneumonia in clinical practice because of its higher prevalence in the community. For similar reasons, influenza virus was the most common cause of the airway-centric pattern (16 cases) compared with RSV (13 cases). Parainfluenza virus, despite its name, is phylogenetically related to RSV and is known to primarily cause bronchitis or bronchiolitis similar to RSV [25]. Our data are limited to reliably determine the imaging appearance of parainfluenza virus infection; however, two of four cases of parainfluenza virus infection presented with a bronchiolitis pattern of disease characterized by bronchial wall thickening and tree-in-bud opacities. Bronchitis, which is caused by both viruses and bacteria, is a clinically important LRTI for which the only imaging manifestation is bronchial wall thickening. The criteria for the diagnosis of bronchial wall thickening in this study were of poor quality. We attempted to measure airway thickness, but the wall thickness of fourth- and fifth-order bronchi is only 1 2 pixels in diameter, introducing excessive stochastic variation in any attempt at objective measurement of wall diameter. As a consequence, we chose to use subjective criteria to determine the presence of bronchial wall thickening. Unfortunately, this finding had the lowest interobserver agreement among the reviewers as measured by the kappa statistic. The development of improved measures of bronchial wall thickening should be an important goal of pulmonary imagers. Improved characterization of bronchial wall thickening would open up new areas for pulmonary investigations. Limitations to this study should be noted. Only a fraction of the total number of patients with viral LRTI underwent CT. This is very likely to have introduced selection bias toward patients with more severe disease. As a consequence, the frequencies of the airway-centric and multifocal pneumonia patterns of disease found in this study are likely to be higher than those found in the overall population with community-acquired viral LRTI. With so many findings being compared across multiple reviewers, there is a possibility that some of the significant differences between categoric variables could be accidental. However, the associations that we discovered have confirmed previously identified characteristics of RSV and adenovirus infections that have been published by other researchers, which would suggest that these associations are true, rather than spurious. There are a variety of other community-acquired viral LRTIs, most notably human metapneumovirus, but also including rhinoviruses, coronaviruses, and herpes viruses. We were unable to evaluated the imaging features of these infections and compare them with those of the viruses for which we did test. We believe that evaluation of the imaging features of these other viruses is an interesting question that should be addressed by further evaluations. Finally, the imaging techniques for the CT examinations were very heterogeneous. This could have resulted in obscuration of subtle differences between viruses. In conclusion, the CT features of bacterial and viral LRTIs overlap considerably, such that, in any individual case, imaging will not reliably differentiate between viral and bacterial infections. RSV and adenovirus tend to have a predilection for distinct patterns of disease. RSV primarily causes an airwaycentric pattern of disease characterized by a variable combination of tree-in-bud opacities, bronchial wall thickening, and peribronchiolar consolidation. Adenovirus primarily causes a multifocal pneumonia characterized by combinations of consolidation and ground-glass opacities in a multifocal distribution without airway findings. Influenza virus has the most variable imaging appearance of the common community-acquired viral infections, with approximately equal frequencies of normal examinations, airwaycentric disease, and multifocal pneumonia. References 1. Elicker BM, Schwartz BS, Liu C, et al. Thoracic CT findings of novel influenza A (H1N1) infection in immunocompromised patients. Emerg Radiol 2010; 17:299 307 2. Shim SS, Kim Y, Ryu YJ. Novel influenza A (H1N1) infection: chest CT findings from 21 cases in Seoul, Korea. Clin Radiol 2011; 66:118 124 3. Mori T, Morii M, Terada K, et al. Clinical characteristics and computed tomography findings in children with 2009 pandemic influenza A (H1N1) viral pneumonia. Scand J Infect Dis 2011; 43:47 54 4. Choi MJ, Lee YS, Lee JY, Lee KS. Novel influenza A (H1N1) virus infection in children: chest radiographic and CT evaluation. Korean J Radiol 2010; 11:656 664 5. Abbo L, Quartin A, Morris M, et al. Pulmonary imaging of pandemic influenza H1N1 infection: relationship between clinical presentation and disease burden on chest radiography and CT. Br J Radiol 2010; 83:645 651 6. Tanaka N, Matsumoto T, Kuramitsu T, et al. High resolution CT findings in community-acquired pneumonia. J Comput Assist Tomogr 1996; 20: 600 608 7. Shiley KT, Deerlin VMV, Wallace T, Miller J, Chest CT. Features of community-acquired respiratory viral infections in adult inpatients with lower respiratory tract infections. J Thorac Imaging 2010; 25:68 75 8. Oikonomou A, Muller NL, Nantel S. Radiographic and high-resolution CT findings of influenza virus pneumonia in patients with hematologic malignancies. AJR 2003; 181:507 511 9. Matar LD, McAdams HP, Palmer SM, et al. Respiratory viral infections in lung transplant recipients: radiologic findings with clinical correlation. Radiology 1999; 213:735 742 10. Marchiori E, Zanetti G, Hochhegger B, et al. High-resolution computed tomography findings from adult patients with Influenza A (H1N1) virus-associated pneumonia. Eur J Radiol 2010; 74:93 98 11. Ko JP, Shepard JA, Sproule MW, et al. CT manifestations of respiratory syncytial virus infection 1094 AJR:197, November 2011

CT of Viral Lower Respiratory Tract Infections in lung transplant recipients. J Comput Assist 16. Escuissato DL, Gasparetto EL, Marchiori E, et al. WC, Graffelman AW, van den Broek PJ, Claas Tomogr 2000; 24:235 241 Pulmonary infections after bone marrow trans- EC. Improved diagnosis of the etiology of com- 12. Gasparetto EL, Escuissato DL, Marchiori E, Ono plantation: high-resolution CT findings in 111 pa- munity-acquired pneumonia with real-time poly- S, Frare e Silva RL, Muller NL. High-resolution tients. AJR 2005; 185:608 615 merase chain reaction. Clin Infect Dis 2005; CT findings of respiratory syncytial virus pneu- 17. Chong S, Lee KS, Kim TS, Chung MJ, Chung 41:345 351 monia after bone marrow transplantation. AJR MP, Han J. Adenovirus pneumonia in adults: ra- 22. Johnstone J, Majumdar SR, Fox JD, Marrie TJ. Vi- 2004; 182:1133 1137 13. Fujita J, Bandoh S, Yamaguchi M, Higa F, Tateyama M. Chest CT findings of influenza virusassociated pneumonia in 12 adult patients. Influenza Other Respi Viruses 2007; 1:183 187 14. Franquet T, Rodriguez S, Martino R, Gimenez A, Salinas T, Hidalgo A. Thin-section CT findings in hematopoietic stem cell transplantation recipients with respiratory virus pneumonia. AJR 2006; 187:1085 1090 15. Ferguson PE, Sorrell TC, Bradstock KF, Carr P, Gilroy NM. Parainfluenza virus type 3 pneumonia in bone marrow transplant recipients: multiple small nodules in high-resolution lung computed tomography scans provide a radiological clue to diagnosis. Clin Infect Dis 2009; 48:905 909 diographic and high-resolution CT findings in five patients. AJR 2006; 186:1288 1293 18. Ajlan AM, Quiney B, Nicolaou S, Müller N. Swine-origin influenza A (H1N1) viral infection: radiographic and CT findings. AJR 2009; 193: 1494 1499 19. Agarwal PP, Cinti S, Kazerooni EA. Chest radiographic and CT findings in novel swine-origin influenza A (H1N1) virus (S-OIV) infection. AJR 2009; 193:1488 1493 20. Oosterheert JJ, van Loon A, Schuurman R, et al. Impact of rapid detection of viral and atypical bacterial pathogens by real-time polymerase chain reaction for patients with lower respiratory tract infection. Clin Infect Dis 2005; 41:1438 1444 21. Templeton KE, Scheltinga SA, van den Eeden ral infection in adults hospitalized with community-acquired pneumonia: prevalence, pathogens, and presentation. Chest 2008; 134:1141 1148 23. Hall C, McCarthey C. Respiratory syncytial virus. In: Mandell GL, Bennett JE, eds. Principles and practice of infectious disease. New York: Elsevier Health Sciences, 2009:2008 2026 24. Shenk T. Adenovirdae: the viruses and their replication. In: Knipe D, Howley P, eds. Field s virology. Philadelphia, PA: Lippincott Williams & Wilkins, 2001:2265 2300 25. Reed G, Jewett PH, Thompson J, Tollefson S, Wright PF. Epidemiology and clinical impact of parainfluenza virus infections in otherwise healthy infants and young children < 5 years old. J Infect Dis 1997; 175:807 813 AJR:197, November 2011 1095