Clinical Severity and CT Features of the COVID-19 Pneumonia: Focus on CT Score and Laboratory Parameters

Background: Although CT characteristics of Coronavirus Disease 2019 (COVID-19) pneumonia between patients with mild and severe forms of the disease have already been reported in the literature, there was little attention to the correlation of imaging features and laboratory testing. We aimed to compare the laboratory and chest CT imaging features in patients with COVID-19 pneumonia between non-severe cases and severe cases, and to analyze the correlation of CT score and laboratory testing. Methods: This study consecutively included 54 patients with COVID-19 pneumonia (26 males and 28 females, 26 to 92 years of age, 43 cases with non-severe and 11 cases with severe group). Clinical, laboratory and image data were collected between two subgroups. A CT score system was used to evaluate the extent of disease. Correlation between the CT score and laboratory data were estimated. Receiver operating characteristic (ROC) analysis was used to evaluate the diagnostic performance of CT score and laboratory tests. Results: Compared with non-severe patients, severe patients had showed increased white blood cell count, neutrophil count, neutrophil percentage, the neutrophil-to-lymphocyte ratio (NLR) and decreased lymphocyte percentage (all p < 0.05). Architectural distortion, pleural effusion, air bronchogram and consolidation-dominant pattern were more common in the severe group (all p < 0.05). CT score of the severe group was higher than the non-severe group (p < 0.001). For distribution characters of the lesions, diffuse pattern in the transverse distribution was more often seen in the severe group (p < 0.001). CT score was positively correlated with the white blood cell counts, neutrophil counts, the percent of neutrophil, NLR, alanine aminotransferase, lactate dehydrogenase and C-reactive protein, and was inversely related to the lymphocyte, the percent of lymphocyte. ROC analysis showed that when the optimal threshold of CT score was 13, the area under the curve was the largest, which was 0.855, and the sensitivity and specicity were 100% and 60% respectively for the diagnosis of the severe patients. Conclusion: CT score showed signicant correlations with laboratory inammatory markers, suggesting that chest [the extent of involvement of reticulation was largest), (6) mixed pattern (combination of GGO, consolidation and reticular opacity) (7) mosaic attenuation, (8) nodules, (9) traction bronchiectasis, (10) architectural distortion, (11) air bronchogram, (12) tree-in-bud, (13) intrathoracic lymphadenopathy and (14) pleural effusion. The detailed descriptions of the imaging interpretation were consistent with the Fleischner guidelines (16). We dened four distributions according to the previous study, including transverse, cranio-caudal, lung region, and scattered (10). Transverse distribution included central type, lesion located in the inner one-thirds of the lung and peripheral type, located in the outer two-third of the lung. Each lung was divided into 3 zones on HRCT as follows in the cranio-caudal plane: upper pulmonary and Lung region lung the CT than the vs. [9.0–17.0], p was a longer time-interval from symptom onset to rst in the severe group (the at 114 days after symptom onset and did not certainly represent irreversible brosis It is not known whether these signs in severe patients with COVID-19 represent irreversible brosis and further follow-up research is needed. Tree-in-bud sign may be due to infectious bronchiolitis in our study. The frequency of tree-in-bud sign is very low, different from H1N1 inuenza A pneumonia which may serve to distinguish SARS-CoV-2 from other types of viral pneumonia. Our research showed that consolidation-dominant pattern and air bronchogram are more common in severe patients, suggesting alveolar exudation is still predominant. There was a high virus load in the body at this time, then alveolar macrophages engulf the virus in large quantities, and lung tissue damage is dominant. Pleural effusion suggested poor prognosis in many studies about COVID-19 14), consistent with our study. Previous studies also showed that the presence of pleural effusion in patients infected with MERS was a poor prognostic indicator For lesion distribution, patients with COVID-19 tended to have peripheral distribution (64.8%), bilateral involvement (98.1%), diffuse distribution in craniocaudal plane (51.9%), and be diffuse in numbers (94.4%), which were consistent with results of previous studies except for lower lung zone distribution The severe patients were usually more diffuse in axial distribution, that is, inferior one-thirds of the lung zone was also affected, indicating that inammation spreads from the periphery to the central lung. CT score was higher in severe patients, suggesting that the affected lungs are more extensive in severe patients, consistent with many studies 14). Our study revealed CT score was associated with a number of laboratory indicators, it’s probably because they're all related to the severity of the disease. The signicance of this result is that if the laboratory indicators change during the follow-up after treatment, it seems that the laboratory indicators can be used instead of frequent CT examination to assess the severity of the disease. The laboratory indicators are more convenient to obtain, and the radiation dose of CT can be reduced correspondingly.


Background
Since December 2019, a cluster of cases of pneumonia with unknown cause has reported in Wuhan, China. On Jan 7, 2020, a novel coronavirus was been identi ed as the causative organism by Chinese scientists (1). On February 11th, the International Committee on  (2). On March 11, 2020, WHO made the assessment that COVID-19 could be characterized as a pandemic. As of April 1, 2020, it has affected 823 626 patients on a globe scale, and causing 40 598 deaths (3).
However, the case fatality rate is very high in some countries, for example, the overall case fatality rate in Italy is about 7.2% (4). So far, China have effectively controlled the epidemic, but unfortunately the virus is rapidly spreading outside of China, especially in America, Italy, Spain, Germany, France, Iran, etc. (3). Typical clinical manifestations of COVID-19 pneumonia include fever, cough, dyspnea (5). In the early phase, typical pathological ndings in the lungs include proteinaceous and brin exudates, mononuclear cells in ltration, desquamation of pneumocytes and hyaline membrane formation (6,7). CT is important in early detection and follow-up after treatment of COVID-19 pneumonia. High resolution computed tomography (HRCT) ndings mainly include peripheral unilateral or bilateral groundglass opacities or consolidation in the early stage (8)(9)(10). In addition, different radiological patterns are observed at different time throughout the disease course (9,11,12). In the currently published literature, the imaging and clinical features of COVID-19 pneumonia between different clinical types have been reported (10,13,14). However, whether there is a correlation between some imaging ndings and laboratory testing has not been reported in the literature. We aimed to describe the clinical characteristics of hospitalized patients with COVID-19 pneumonia and compare the laboratory tests and chest CT imaging features between severe cases and non-severe cases and to analyze their correlation between the imaging features and laboratory tests.

Patients
The Institutional Review Board granted approval for our retrospective study, waiving informed consent because of its retrospective nature. This study consecutively included 54 patients with COVID-19 from February 4, 2020 to March 23, 2020. All available clinical, laboratory and imaging data were collected for all patients. Two patients with underlying lung diseases (tuberculosis and small cell lung cancer) were excluded, because the evaluation of the extent of the lesion may be affected. Forty-eight cases were diagnosed by real time reverse-transcription-polymerase chain-reaction (RT-PCR), six cases were clinically diagnosed, according to the guideline on COVID-19 (trial version 5) (15) issued by the China's National Health Commission. According to diagnosis and treatment of COVID-19 (trial version 5) (15), patients were classi ed into common, severe and critical type in our study, and common type were included in the nonsevere group, while severe and critical type were included into the severe group. All patients underwent CT scan after admission.

Chest Ct Protocol
All chest CT examinations were performed on a third-generation dual-source CT system (SOMATOM Force; Siemens Healthineers, Germany). Images were acquired in the supine position after end-inspiration and extended from the lung apices to the costophrenic angles by using the following parameters: tube voltage, 120 kVp; automated tube current modulation (Care Dose 4D); thin collimation, 0.75 mm; pitch, 0.6; rotation time, 500 ms; reconstruction matrix, 512 × 512. CT images were reconstructed into 1.0-mm section thickness by using a high spatial revolution algorithm. All CT scans were obtained with window settings that were appropriate for lung parenchyma (window width, 1300 Houns eld units [HU]; level, -450 HU) and mediastinum (window width, 400 HU; level, 40 HU).

Image Evaluation
The thin-section CT images were assessed in totally random order by two radiologists (with 6 years and 11 years of experience in chest CT imaging, respectively) without reference to the clinical or laboratory test results. Any disagreement was resolved by discussion and consensus. We evaluated 14 imaging features, including (1) ground-glass opacities (GGO), (2) consolidation, (3) crazy-paving sign, (4) reticulation, (5) the prominent pattern of opacities (according to the extent of involvement, we divided the prominent pattern into GGOdominant type [the extent of involvement of GGO was largest], consolidation-dominant type [the extent of involvement of consolidation was largest] and reticulation-dominant type [the extent of involvement of reticulation was largest), (6) mixed pattern (combination of GGO, consolidation and reticular opacity) (7) mosaic attenuation, (8) nodules, (9) traction bronchiectasis, (10) architectural distortion, (11) air bronchogram, (12) tree-in-bud, (13) intrathoracic lymphadenopathy and (14) pleural effusion. The detailed descriptions of the imaging interpretation were consistent with the Fleischner guidelines (16). We de ned four distributions according to the previous study, including transverse, cranio-caudal, lung region, and scattered (10). Transverse distribution included central type, lesion located in the inner one-thirds of the lung and peripheral type, located in the outer two-third of the lung. Each lung was divided into 3 zones on HRCT as follows in the cranio-caudal plane: the upper zone above the carina; the lower zone below the inferior pulmonary veins; the middle zone between the upper and lower zone. Lung region was categorized as unilateral or bilateral lung involvement. Scattered distribution was categorized as focal (one lesion), multifocal (two or more lesions, localized in two lobes), diffuse (multiple lesions, in more than two lobes). A CT score system was used to evaluate the extent of disease following the criteria de ned in the previous study (11). A thinsection CT score was assigned on the basis of all abnormal areas involved. A semi-quantitative visual score system was assigned for each lobe based on the following: score 0, 0% involvement; score 1, less than 5% involvement; score 2, 5-25% involvement; score 3, 26-49% involvement; score 4, 50-75% involvement; and score 5, greater than 75% involvement. There was a score of 0-5 for each lobe, with a maximal score was 25.

Statistical analysis
All data were expressed as mean ± standard deviation (SD), n (%) or median ± interquartile range (IQR). Continuous variables between the non-severe and the severe two subgroups were compared by using Mann-Whitney U test, and categorical variables were performed by χ2 or Fisher's exact test. The correlations between CT score and laboratory tests were evaluated by Spearman correlation analysis. The performance of laboratory tests of obvious differences between the two groups and CT score were evaluated by receiver operating characteristic analysis. Then all statistical analyses were performed using a statistical software package (SPSS 23.0, Inc., Chicago, IL, USA) and MedCalc software (version 15.2.2, MedCalc Software). Two-sided p values < 0.05 was considered to be statistically signi cant.

Page 4/15
The 54 patients included 26 males and 28 females, 26 to 92 years of age (mean age 60 ± 14 years). The demographics and clinical characteristic are shown in Table 1. In the full cohort, the mean age was 60 years (SD 14; range of 26-92), including 26 (48.1%) men and 28 (51.9%) female. Epidemiological history was divided into three types. All patients were from epidemic area, 53 of them were from Wuhan, one was from Xiaogan. Four patients (7.4%) had direct exposure history to patients with con rmed disease, and 13 patients (24.1%) had a disease associated with a family outbreak. Half of patients had one or more underlying diseases, including diabetes (16 [ . Less common symptoms were chest pain, sore throat and headache ( Table 1). As of the end of the study, 45 patients were discharged, six remained in the hospital, three were transferred to other hospital and no patient died. The median time-interval from symptoms onset to rst CT examination was 19 day. The median time-interval between rst CT examination and laboratory testing was 2 day. Comparison of basic clinical characteristics, laboratory tests and CT ndings in the non-severe and the severe groups Among the 54 included patients, 43 (20 men, 23 women) were non-severe group, and 11 (six men, ve women) were severe group. No signi cant differences in age (p = 0.186), gender (p = 0.634), and underlying disease (p = 0.735) were found between the two groups.
Patients of severe group showed increased white blood cell (WBC) count (9.1 × 10 9 /L), neutrophil count (7.9 × 10 9 /L), neutrophil  Table 2). There was no obvious difference in levels of lymphocyte count, the percent of monocyte, aspartate aminotransferase (AST) and CRP between the two groups (all p > 0.05), while levels of LDH had the tendency for increase in severe group.
The percent of monocyte (%) 8.     1 and 2). Only one patient had tree-in-bud sign and pneumothorax in our study (Fig. 2B). Four of the 14 imaging features -architectural distortion ( Fig. 2A), pleural effusion ( Fig. 2A), air bronchogram (Fig. 3A) and the prominent pattern of opacities -were statistically signi cant between the two groups (all p < 0.05) ( Table 2). There was no obvious difference in reticulation, traction bronchiectasis, GGO, consolidation, crazy-paving sign, mixed pattern, mosaic attenuation, nodules, tree-in-bud, and intrathoracic lymphadenopathy between the two groups (all p > 0.05). The distribution pattern lesions in two groups are also shown in Table 2 Correlation between CT score and laboratory tests and their performances for severe patients of COVID-19 pneumonia CT score was positively correlated with the cell counts of WBC, neutrophil, the percent of neutrophil, NLR, ALT, LDH and CRP, and inversely related to the lymphocyte, the percent of lymphocyte. CT score was not correlated with monocyte count, monocyte percentage and AST (Table 3). The performance of CT score and laboratory tests for severe patients of COVID-19 pneumonia are also shown in Table 4. The area under the curve of CT score, the percent of lymphocyte, neutrophil count, NLR, the percent of neutrophil and WBC were 0.855, 0.828, 0.827, 0.820, 0.789 and 0.786 respectively. The CT score demonstrated the best performance when the cut-off value of CT score was 13.

Discussion
It has been reported that about 70%-85% of COVID-19 patients have a mild form of disease, a self-limiting disease, while approximately 15%-30% of patients have a severe form of disease [17][18][19]. Most of mild patients can heal themselves, while severe patients need medical intervention. If some clinical or imaging indicators can differentiate them, it will help clinicians choose more bene cial treatment regimens. In this study, 20.4% of patients were severe after admission to the hospital, receiving ICU care, while 79.6% of patients were non-severe, receiving general therapy (i.e. supportive therapy, oxygen therapy, antiviral and antibiotic therapy). Our results indicated that some laboratory and imaging indicators of COVID-19 patients in severe group can be distinguished from those in nonsevere group. And we semi-quantitatively evaluated the extent of involvement of the lesion and found that it was related to many indicators of laboratory testing.
In our study, most hospitalized patients with COVID-19 pneumonia were elderly. There was no obvious predilection for sex and underlying disease in full cohort, consistent with previous reports (17). The epidemiological history showed that patients with a close contact accounted for only 7.4%, while patients with familial aggregation accounted for 24.1%. It suggested that the con rmed patients should be strictly isolated from their families, and the isolation measures similar to the shelter hospital may curb the spread of the epidemic. Patients with underlying diseases accounted for half of the cases, higher than reported literatures (5,18), which may be related to the fact that most of the patients in our cohort were elderly. Fever, cough, sputum production, dyspnea and short of breath were the most common symptoms in patients with COVID 19 pneumonia, indicating that the receptor of target cells of might be primarily located in the lower airway, consistent with reported literature (5,17,18,20). However, the intestinal symptoms such as diarrhea seemed more prominent in our cohort than previously reported, which may be related to the difference in the route of infection and the course of disease.
Our results showed that there was no obvious difference in age, gender and underlying diseases between the two groups, but previous studies showed that patients with severe disease were generally older, and had a greater number of underlying diseases than those with non-severe disease [17,18]. These results may be related to the insu cient sample size. Compared with non-severe group, some laboratory indicators such as the leukocyte count, neutrophil count, NLR and the percentage of neutrophils were higher in severe patients, which may indicate that severe patients may be associated with other microbial infections, such as bacterial infection. In addition, neutrophilia may be related to cytokine storm induced by virus invasion. The percentage of lymphocytes was lower in severe patients, which may be due to the increased proportion of neutrophils. The lymphocyte count of the two groups decreased, but there was no signi cant difference between two groups. In general, the lymphocyte count of viral infection increased, but SARS-CoV-2 infection, similar to severe acute respiratory syndrome (SARS) (21,22) and the Middle East respiratory syndrome (MERS) (23,24), also caused the lymphopenia of peripheral blood. In the published literature, most patients had the normal ALT and AST and increased LDH in the whole cohort, consistent with previous reports (17,18). However, the levels of ALT, AST and LDH of severe patients were higher than those of mild patients, which were inconsistent with our studies. Some researchers believed that virally induced cytotoxic T cells and the induction of a dysregulated innate immune response were a more probable explanation for the association between deranged liver markers and COVID-19 disease severity (25).
Chest CT has a high sensitivity for diagnosis of COVID-19 pneumonia (25). It can be used in the early screening of highly suspected cases, especially for the patients whose PCR are negative for the rst time. Our study showed ground-glass opacities (100%), consolidation (92.6%), reticulation (66.7%), traction bronchiectasis (57.4%), mixed pattern (57.4%) and architectural distortion (53.7%) are the most common imaging ndings. These features may correspond to the alveolar exudation combined with broblasts proliferated in the alveolar cavity and interstitium. It suggests that the in ammatory process of injury and repair occur simultaneously in the lesions. However, unlike other published studies (13,14,27), our study showed that the frequency of consolidation and reticulation is higher. This may be related to the long course of infection leading to organic pneumonia. That is, with the increase of time course, the frequency of reticulation and consolidation increase in our study. This study showed that architectural distortion was more likely to be seen in severe patients, and traction bronchiectasis has a high frequency in severe patients. These ndings may indicate pulmonary brosis. A follow-up study of patients recovering from SARS showed that traction bronchiectasis was recoverable at 114 days after symptom onset and did not certainly represent irreversible brosis (28). It is not known whether these signs in severe patients with COVID-19 represent irreversible brosis and further follow-up research is needed. Tree-in-bud sign may be due to infectious bronchiolitis in our study. The frequency of tree-in-bud sign is very low, different from H1N1 in uenza A pneumonia (29), which may serve to distinguish SARS-CoV-2 from other types of viral pneumonia. Our research showed that consolidation-dominant pattern and air bronchogram are more common in severe patients, suggesting alveolar exudation is still predominant. There was a high virus load in the body at this time, then alveolar macrophages engulf the virus in large quantities, and lung tissue damage is dominant. Pleural effusion suggested poor prognosis in many studies about COVID-19 (10,14), consistent with our study. Previous studies also showed that the presence of pleural effusion in patients infected with MERS was a poor prognostic indicator (30). For lesion distribution, patients with COVID-19 tended to have peripheral distribution (64.8%), bilateral involvement (98.1%), diffuse distribution in craniocaudal plane (51.9%), and be diffuse in numbers (94.4%), which were consistent with results of previous studies except for lower lung zone distribution (10). The severe patients were usually more diffuse in axial distribution, that is, inferior one-thirds of the lung zone was also affected, indicating that in ammation spreads from the periphery to the central lung. CT score was higher in severe patients, suggesting that the affected lungs are more extensive in severe patients, consistent with many studies (10,14). Our study revealed CT score was associated with a number of laboratory indicators, it's probably because they're all related to the severity of the disease. The signi cance of this result is that if the laboratory indicators change during the follow-up after treatment, it seems that the laboratory indicators can be used instead of frequent CT examination to assess the severity of the disease. The laboratory indicators are more convenient to obtain, and the radiation dose of CT can be reduced correspondingly.
There are some limitations in our study. First, due to the small sample size of severe patients, it may cause selection bias. Second, the time-interval between laboratory testing and the CT examination was variable, which may limit the ability to determine the relationship between laboratory tests and CT score. However, the median time-interval between laboratory testing and the CT scan was 4 day, during this time interval, it may not cause major changes in indicators. Third, because the diagnosis standard of severe pneumonia is not uniform, it can cause reference test bias. Finally, CT scoring is a semi-quantitative scoring system, which still has some subjectivity. Arti cial intelligence is expected to play a more important role in future research.

Conclusion
In conclusion, CT score showed signi cant correlations with laboratory in ammatory markers, suggesting that chest CT and laboratory examination maybe provide a better reference for clinicians to judge the severity of diseases.

Consent for publication
Not applicable.

Competing interests
We declare that we have no known competing nancial interests or personal relationships that could have appeared to in uence the work reported in this paper.

Funding
None.

Authors' contributions
All authors meet the author requirements according to ICMJE guidelines. All authors read and approved the nal manuscript.  COVID-19 pneumonia in a 42-year-old man who presented with fever, cough for 47 days and dyspnea for 39 days before CT scan ultimately requiring the treatment of extracorporeal membrane oxygenation (ECMO). Chest CT scan con rmed with severe COVID-19. A. Axial thin-section unenhanced CT scan demonstrates diffuse ground-glass opacities, patchy consolidation, prominent traction bronchiectasis and architectural distortion (white arrows) in bilateral upper lobe and superior segment of bilateral lower lobe, and demonstrates a small amount of right pleural effusion and interlobular effusion (black arrows). B. Axial unenhanced image at the bifurcation level of intermediate bronchus shows that diffuse ground-glass opacities, patchy consolidation and tree-in-bud sign (white arrow), and shows bilateral pneumothorax (black arrows). CT extent score is 25. COVID-19 pneumonia in a 62-year-old woman who presented with fever, cough for 26 days and dyspnea for 16 days before CT scan.
Chest CT scan con rmed with severe COVID-19. On unenhanced CT scan, subpleural consolidation and air bronchogram (arrow) can be seen in the superior segment of the bilateral lower lobes, and scattered consolidation and GGO are distributed along the bronchovascular bundle or subpleural area of both lungs. The prominent pattern of opacities is consolidation.