Performance of Chest Computed Tomography in Differentiating Coronavirus Disease 2019 From Other Viral Infections Using a Standardized Classification
Borges da Silva Teles, Gustavo, Kaiser Ururahy Nunes Fonseca, Eduardo, Yokoo, Patricia, et al.
J Thorac Imaging. 2021;36(1):31-36. doi:10.1097/RTI.0000000000000563.
Radiologists from the University of Sao Paulo report interesting outcomes from a retrospective study, wherein two thoracic radiologists evaluated the performance of a classification system in detecting COVID-19 in a patient population based on the “Radiological Society of North America’s Statement on Reporting Chest CT Findings Related to COVID-19″.
An original cohort of 350 patients who were imaged due to concerns associated with infection with the novel coronavirus at their university between March 15th and 24th 2020. Patients without COVID-19 RT-PCR results, patients with positive COVID-19 RT-PCR and respiratory pathogen panels (RPPs) and patients with negative COVID-19 RT-PCR and RPPs were excluded from the trial rounding out the final cohort to 175. Patients’ demographic information, comorbid diseases, date of CT scan and date of initial symptoms were recorded. All patients who were imaged underwent non-contrast-enhanced CT scans at end expiration with 1mm reformations. Radiologists who were blinded to the viral panels’ results reviewed these retrospective studies and assigned each study with one of four different categories of findings based on the aforementioned RSNA statement; including “typical”, “atypical”, “indeterminate” and “negative”. Two different scenarios were instituted for quantitative assessment. Scenario 1 utilized only scans labeled as “typical” as positive results, while all scans labeled “indeterminate”, “atypical” and “negative” were categorized as negative studies. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy were then calculated. Scenario 2 counted tests marked as “typical” and “indeterminate” as positive studies, while “atypical” and “negative” scans were again classified as negative studies. The same quantitative measures were applied for comparison.
Of the 175 patients included in the final cohort, 87 patients had a positive COVID-19 RT-PCR test and 88 patients had a negative COVID-19 RT-PCR test and a positive RPP. 64 of the 87 positive patients (73.6%) had CT scans that were classified as having “typical” findings for COVID-19, while only 2/88 patients (2.3%) were classified as typical (p<.001). The COVID negative group had scans classified as “negative” 60.2% while the COVID positive group had 14.1% ascribed as “negative” (p<.001). Inter-reader agreement between the two radiologists was good-excellent with a k=.80 (.73-.87), suggesting reliability of the radiologists involved. Scenario 1, whereby only CT scans classified as having findings “typical” for COVID were considered positives yielded a sensitivity of 73.6% (CI 95%: 63-82.4%), specificity of 97.7% (CI 95%: 92-99.7%), PPV of 97% (CI 95%: 89.5-99.6%), NPV of 78.9% (CI 95%: 70-86.1%) and an accuracy of 85.7% (CI 95%: 79.8-85.1%). Scenario 2, which positive studies were characterized as having either “typical” or “indeterminate” findings yielded a sensitivity of 82.8% (CI 95%: 73-90%), specificity of 87.5% (CI 95%: 78.7-93.6%), PPV of 86.7% (CI 95%: 77.5-93.2%), NPV of 83.6% (CI 95%: 74.5-90.6%) and an accuracy of 85.1% (CI 95%: 79.8-85.1%).
There were significant differences in the sensitivity, specificity, PPV observed between scenario 1 and 2, with scenario 1 outperforming 2. Significant differences observed between the “typical” and “negative” scans in the COVID positive and negative groups were also present. There was a noticeable amount of false negative studies in the COVID-19 positive cohort. The thought behind the number of “negative” studies in the positive group is likely related to the number of people presenting in the early phase of disease (<5 days from symptom onset), which accounted for 99.9% of the negative studies (only one patient presenting in the late phase had a “negative” study). This study shows that a CT scan with “typical” findings for COVID-19 pneumonia can be relatively sensitive and highly specific for infection with the novel coronavirus, and that these findings have a high level of inter-observer reliability. This study had a few limitations. Mainly, the small sample size from a single academic institution could lead to sampling bias. Also, given the retrospective nature of this study and the fact that this study was conducted during a global pandemic, the radiologists may have a higher level of internal sensitivity/specificity. Regardless, this trial seems statistically sound; and although the majority of medical societies do not condone the use of CT as a screening modality, in the right patient population CT can be an invaluable tool for the diagnosis of COVID-19 pneumonia.
Fleischner Society Visual Emphysema CT Patterns Help Predict Progression of Emphysema in Current and Former Smokers: Results from the COPDGene Study
El Kaddouri, Bilal., Strand, Matthew J., Baraghoshi, David., Humphries, Stephen M., Charbonnier, Jean-Paul., Van Rikxoort, Eva M., Lynch, David A.
Radiology. 2021; 298(2): 441-449. doi: 10.1148/radiol.2020200563.
In an interesting article published by a multinational group of researchers from Belgium, the Netherlands, and the United states, researchers described visual patterns of emphysema and their association with patients’ progressive air-trapping and functional status during follow up. This study was a cohort trial and included current and former smokers with and without chronic obstructive pulmonary disease (COPD) who were enrolled in the prospective Genetic Epidemiology of COPD study (COPDGene). Of the original 4995 patients who had completed a second visit within five years of their first, 829 patients were excluded due to their change in smoking status or their lack of requisite imaging, leaving 4166 patients in the cohort. 1655 patients (roughly 41%) carried a formal diagnosis of COPD by spirometric assessment. CT scans were obtained per the COPDGene study parameters at both full inspiration and passive expiration and reconstructed to sub-millimeter slice thickness using a medium sharp resolution algorithm. A 3D sliver software was used to perform quantitative analysis on the extent of emphysema present in each study. Radiologists were tasked with categorizing emphysema as either centrilobular (CLE) or paraseptal (PSE) and grading the emphysema using the updated Fleischner guidelines (for CLE: trace, mild, moderate, confluent or advanced destructive (AD) emphysema; for PSE: mild or substantial). Patients’ demographic information, bronchodilator responsiveness, 6-minute walk score, dyspnea score and whether or not patients carried a diagnosis of COPD were utilized in data collection.
Given that CLE and PSE patterns of emphysema often coincide, it was left to the Radiologists’ discretion of what the predominant pattern the patient was exhibiting. A third category of “mixed” emphysema was proposed, however researchers found too many confounders within this group to be statistically feasible. For patients with visible trace, mild and moderate CLE on baseline CT, researcher’s observed increased rates of progression of emphysema and air trapping. Increased severity of CLE was also associated with older age (avg of 59 years for trace CLE and avg of 66 years for AD CLE; P<.001), lower weight (86kg for trace CLE and 75kg for AD CLE), non-Hispanic white race, longer smoking history (40 pack-years for trace to 58 pack years AD CLE, P<.001), and a lower prevalence of current smokers (52% current smokers with trace, 11% current smokers with AD CLE, P<.001). PSE patterns of emphysema were observed in 1010 of 4166 patients (24%); 58% were graded as mild and 42% were graded as substantial. Visual presence of PSE on CT was observed more frequently in non-white participants and more persistent current smokers, a longer smoking history (39.1 years for absent, 45.2 years for mild, and 49.3 for substantial, P<.001). Greater severity of PSE was related to lower weight (85kg for mild and 81kg for substantial, P<.001) and male sex (56% men in mild population, 69% men in substantial). In both CLE and PSE more advanced patterns of disease were associated with increased GOLD stage (CLE 0% pts with trace to 29% with GOLD 4 in AD CLE, P<.001) (PSE 1% of pts with mild to 3% with GOLD 4 in substantial PSE, P<.001), worsening airflow obstruction (CLE from 84% predicted FEV1 in trace group to 46% predicted FEV1 in AD CLE, P<.001) (PSE 81% predicted FEV1 in mild to 76% predicted FEV1 in substantial, P<.001), and decreased distance achieved during the 6-minute walk test (trace CLE with 1441 feet to 1198 feet in AD CLE, P<.001) (PSE mild 1419 feet to 1394 feet in substantial PSE).
Smokers who were diagnosed with COPD and who also exhibited visual emphysema on baseline CT scan showed a decrease in lung density of –5.1g/L (95% CI –6.0, -4.1; P<.001) a significant decrease compared to those without baseline emphysema who exhibited a decrease in lung density of -.1 g/L (95% CI –1.4, 1.3; P=.92). African American participants showed larger decreases in lung density compared to their White counterparts, exhibiting a decrease of 6.7 g/L (95% CI 5.5, 8.0) compared to 4.6 g/L (95% CI 3.7, 5.6); P<.001. Decreases in lung density showed predictable associations with worsening emphysema, except in the AD CLE group. This is thought to be explained by the degree of confluent destructive change associated with late-stage disease. The presence of paraseptal emphysema was associated with increased progression of air trapping, which worsened as the PSE became more severe. The theory behind this is related to increased ability of cystic changes at the periphery of the lung to expand more freely to form bullae, which can cause compression on the more normal areas of the lung. It was also independently noted that smokers with PSE also had increased numbers of pack years when compared to their CLE counterparts. This study nicely outlines how the presence of visual emphysema is an independent and reliable predictor of progressive airway obstruction in smokers or former smokers with or without COPD. This study has a few limitations. A major limitation is the binary race classification system used; that being non-Hispanic-White or non-Hispanic-Black. Not including patients of races other than these two likely limits the information potentially gained from a study of this magnitude; especially given the race-related decreases in lung density and air-trapping observed in this study. Secondly, since CLE and PSE often coincide, it would be advantageous to the researchers and to the patient-population to formulate a way to include this mixed phenotype group such that more real-world scenarios would be more thoroughly investigated. It would also be interesting to see how the participants who were excluded due to changes in their smoking status faired during the intervening five years.
Social Distancing with Portable Chest Radiographs During the COVID-19 Pandemic: Assessment of Radiograph Technique and Image Quality Obtained at Six Feet and Through Glass
Christopher P. Gange, Jay K. Pahade, Isabel Cortopassi, Anna S. Bader, Jamal Bokhari, Matthew Hoerner, Kelly M. Thomas, Ami N. Rubinowitz
Radiology: Cardiothoracic Imaging; Volume 2, Issue 6
Researchers from Yale University sought to investigate the utility of allowing portable chest radiographs to be taken through a glass door, in order to reduce exposure to potential COVID-19 patients while obtaining diagnostic quality imaging. Due to the need to don personal protective equipment (PPE) each time a technologist was to take a portable radiograph, any possible way to reduce using PPE is encouraged as there are shortages across the world. The researchers theorized that a portable radiograph could be obtained through glass doors in their emergency department to help reduce potential exposure to the virus, reduce PPE usage, and provide valuable diagnostic information.
There were two techniques used: a standard technique; and a modified technique. Standard technique included full donning of PPE, the x-ray unit was brought into the patient’s room and placed at the foot of the bed, and detector placed behind patients back with a distance of roughly 50 inches between machine and detector. The modified technique included a distance of 72 inches which allowed the technologist and x-ray unit to be outside of the room with the detector placed by a nurse already donned in PPE in the patient’s room. The modified technique required higher quantity and energy of radiation to have enough penetration through the door and low exposure time to reduce motion blur. SmartGrid software was used to adjust the increased scatter radiation to help with image quality than the standard conventional anti-scatter grid. Exposure index values allowed for estimation of radiation exposure of patients at the detector; EI value chosen for the modified technique was 300 kVp rather than the 200 kVp for standard but the true value for the individual radiograph is varied and based on the patient’s BMI. For comparison between the standard technique and modified technique, 50 radiographs of each technique were randomized and evaluated by three thoracic radiologists. Images were rated as diagnostic or nondiagnostic, as well as any parenchymal abnormalities noted.
The modified technique resulted in higher EI to the detector across all patients (p <0.001). Patients BMI had a role in the entrance skin exposures, with patients with a >50th percentile BMI had a higher exposure index with the modified compared to standard technique. Image quality was rated as diagnostic for each of the 100 radiographs for two of the radiologists, while the third radiologist said that 3 studies were not diagnostic (2 were standard technique, 1 was modified). Technologists surveyed who had used the new technique felt safer, recognized the decrease use of PPE, and felt that the modified technique was as easy if not easier to perform. Overall, this study showed a good initial set of data where this modified protocol could be useful for keeping the radiology team safe while also providing diagnostic level information.
Limitations for this study include a small sample size and there was no comparison from modified technique to a standard technique on the same patient. Additionally, this technique is environment dependent as not all emergency rooms have a glass door for a patient’s room. This study at least provides a framework that individual institutions could use to develop their own protocol.
Multimodality Assessment of Thoracic Aortic Dimensions: Comparison of CTA, MRI and Echocardiography Measurements
Caio Frazao, Anahita Tavoosi, Bernd J Wintersperger, Elsie T Nguyen, Rachel M Wald, Maral Ouzounian, Kate Hanneman
Journal of Thoracic Imaging, April 3 2020, online ahead of print edition
Researchers from the Toronto General Hospital sought to evaluate intermodality differences of thoracic aortic measurements using CTA, MRI, and echocardiography. These three modalities are used for surveillance in patients that have known or suspected thoracic aorta aneurysms and the modality chosen is typically due to a variety of factors including clinical status, diagnostic question, local expertise, and availability. There is wide variability as far as protocol goes for measuring the aorta between different modalities and different organizations and the purpose of this study was to compare intermodality differences as well as the different techniques for measurement.
This was a retrospective study that selected 127 patients that had undergone CTA and MRI evaluation of the aortic root and/or thoracic aorta within 6 months of each other. 23 of these patients had an available transthoracic echocardiography report which was included in analysis. All patients included had the scans within at least 72 days of each other. Exclusion criteria included interval cardiac surgery or aortic dissection. CTA was acquired with 64 or 320 detector systems under a variety of protocols, but typical scan parameters included 0.5 mm collimation, 0.5-5 mm reconstructed slice thickness, 100-120 kVp, and tube current modulation according to body habitus. MRI was performed on a 1.5 or 3T Siemens scanner with retrospectively gated SSFP images through the aortic root. MRA images were obtained via non-ECG gated contrast enhanced MRA. Measurements were obtained by a thoracic trained radiologist at the sinuses of valsalva, sinotubular junction (STJ), proximal ascending aorta (1 cm from STJ), distal ascending aorta (defined as 1 cm below the origin of the first aortic arch branch), maximum diameter ascending aorta, proximal aortic arch, distal aortic arch, distal descending aorta, and maximum descending aorta. STATA software was used for statistical analysis.
Inter-lumen to inter-lumen (II) or outer-lumen to outer-lumen (OO) measurements were used. With both of these measurements, CT and MRI did not show differences at the aortic root. The II measurement did show a difference at the distal descending aorta (20.2 vs 19.8, P <0.001). TTE aortic root diameter was significantly smaller than CT using the OO technique, as well as the II technique (mean difference 4.9 mm, P <0.001). No differences were seen between TTE and CT were seen of the proximal ascending aorta. TTE aortic root diameter was significantly smaller than MRI using the II and OO technique ( mean difference 4.8 mm; P <0.001). There were no significant differences between the TTE and MR measurements of the proximal ascending aorta. Overall, these results are important for showing that CT or MR could be useful for evaluating the aorta and there is excellent agreement between the two modalities. In contrast, TTE shows a significantly smaller diameter than both CT and MR at the aortic root. Overall, excellent agreement between CT and MR aortic measurements can help guide modality choice for future surveillance of aortic aneurysms and radiologists should state which measurement technique (II or OO) they used to ensure reproducibility/consistency for further studies.
There are multiple limitations to this study. First, there are variations in scanning protocols due to the retrospective nature of the study. Second, the patients did not have all three modalities which limits the power of the study. Third, studies were performed during diastole so measurements between the different cardiac phases could provide differences in measurement. Finally, theoretically the time length between studies could pose an issue as there could be changes in aortic morphology in that time frame.
References
