The added value of chest high-resolution CT ndings of Corona Virus Disease 2019 in predicting severity of disease

The aim of this study was to retrospectively analyze chest thin-section high-resolution CT (HRCT) ndings for 32 patients with Corona Virus Disease 2019 (COVID-19) and clarify the correlation between CT data and laboratory results. 30 patients presented with abnormal initial CT scans. Of 30 patients, COVID-19 showed the involvement of bilateral lungs in 24 (80%), involvement of more than two lobes in 24 (80%), ground-glass opacities without consolidation in 27 (90%), ground-glass opacities with consolidation in 23 (76.7%), opacities with irregular intralobular lines in 26 (86.7%), opacities with round morphology in 25 (83.3%), and peripheral distribution in 30 (100%). Pleural effusion or mediastinal lymphadenopathy was relatively rare manifestations. Rapidly progression of the disease demonstrated by increasing number and range of ground glass opacities and appearance of consolidations at follow-up CT images in two patients. The CT lung severity score and No. of lobes involved were negatively correlated with lymphocyte count(r=-0.363, P=0.041; r=-0.367, P=0.039, respectively). Chest HRCT of COVID-19 predominantly manifests multiple, round, ground glass opacities with irregular intralobular lines, and peripheral distribution of bilateral lungs. HRCT is a potential tool for early screening, assessing progress, and predicting disease severity of COVID-19. Authors Jie Zhou and Jie Cao contributed equally to this work and are co-rst authors.

i.e. 2019 novel coronavirus (2019-nCoV)3. The pathogen was classi ed in the beta genus4 and rst isolated in the Wuhan seafood market5. Like the severe acute respiratory syndrome (SARS), the humanto-human transmission among persons exposed to COVID-19 infection has been documented recently6,7. The inhalation droplets or aerosol transmission and contact transmission of COVID-19 have been con rmed; and faecal-oral transmission has yet to be further con rmed7. The incubation period of COVID-19 ranges from 1 to 14 days, usually from 3 to 7 days8. Therefore, the 14-day medical observation or quarantine period would be the best management for exposed persons. With the spread of this coronavirus, COVID-19 has been found in many cities of China and abroad9. International coordination and cooperation will be essential to control the spread of the COVID-19 and to prevent the explosive super-transmission events.
Based on current experience, pulmonary imaging appearances occur earlier than clinical symptoms10.
Patchy/punctate ground glass opacities are the most common manifestations of COVID-19 pneumonia10-12. Chest digital radiography has a limited effect in showing ground glass opacities due to its overlapping imaging characteristics. Similar to the CT ndings of SARS11, ground glass opacities of COVID-19 may be accompanied by interlobular septal thickening and irregular intralobular lines.
These imaging changes manifest pathologically by interstitial and intra-alveolar edema, mild interstitial in ltration with in ammatory cells, and vascular congestion13. High resolution CT (HRCT) is the most sensitive imaging technique for exhibiting early lung changes such as punctate ground glass opacities14.
Thin-section CT imaging is more sensitive to exhibit interlobular and intralobular septal changes than thick-slice CT images and provides more detailed radiological features15.
To the best of our knowledge, there are a few reports focused on the imaging features of the COVID-19 10,11. Therefore, we analyzed the HRCT imaging ndings of 32 patients with COVID-19 in this study.
The aim of present study was to demonstrate the pulmonary characteristic manifestations of COVID-19 for early identi cation of this communicable disease and early isolation of patients, which enabled early implementation of public health surveillance, control and response to the epidemic.

Results
Clinical and pathological characteristics of patients 20 males and 12 females were included in the study (age ranged from 19 to 80 years; mean age, 49.09±16.89 years). 10 patients had a history of residence in or travel to Wuhan, while 17 patients had exposed to infected patients of COVID-19. Some of them were family cluster cases. The clinical symptoms of COVID-19 included fever (84.38%), cough (59.38%), fatigue (15.63%), headache (18.75%), muscle soreness (9.38%), and nausea (15.63%). The clinical characteristics of patients were summarized in Table 1. White blood cell counts of most COVID-19 patients (84.4%) were normal and the rest (15.6%) were lower. Lymphocyte counts of 16 COVID-19 patients (62.5%) were lower and the rest (37.5%) were normal. Neutrophil counts of most COVID-19 patients (93.8%) were normal and the rest (6.2%) were higher or lower. The laboratory characteristics of patients were summarized in Table 2.

Follow-up Chest CT
Three patients underwent follow-up chest CT scan during the study date range, one of whom performed two follow-up CT scans. The time interval between initial CT and follow-up CT scans was 2 days. The initial CT and follow-up CT revealed no pulmonary lesions in 1 patient. The other 2 patients showed rapidly progression that manifested by increasing number and range of ground glass opacities and pulmonary lesions (Fig. 3).

Correlation analyses between the CT data and laboratory data
The number of lobes involved showed a statistically negative correlation with lymphocyte count (r = -0.367, P = 0.039) ( Table 4, Fig. 4a). The CT lung severity score was negatively correlated with lymphocyte count (r = -0.363, P = 0.041) ( Table 4, Fig. 4b). No signi cant correlation was found between the number of lobes involved or the CT lung severity score and white blood cell count (r = -0.197, P = 0.281; r = -0.135, P = 0.460; respectively) ( Table 4).

Discussion
According to the guideline revised by National Health Commission of China (Trial version 6), the con rmed diagnosis of COVID-19 should be based on the positive 2019-nCoV detection. However, the positive imaging manifestation is one of the indispensable clinical diagnostic criteria. Moreover, the sensitivity of 2019-nCoV nucleic acid detection is poor despite its high speci city16. Therefore, it is of great signi cance to accurately recognize the imaging features of COVID-19 for its rapid screening and early diagnosis. Chest HRCT with thin-section is currently considered to be one of the most effective tools to early screening and accurate assessment for COVID-19 owing to its high sensitivity and convenience.
2019-nCoV is highly homologous to the previous SARS-CoV2,17. The pathological features of COVID-19 are greatly similar to those seen in SARS and Middle Eastern respiratory syndrome (MERS) coronavirus infection18,19. Type alveolar epithelium is the target cell of the coronavirus. Similar to SARS CoV, 2019-nCoV adheres rstly to alveolar epithelium in peripheral lobules and then damages alveolar walls causing interstitial and intra-alveolar edema, interstitial in ammatory in ltration, dominated by lymphocytes13,20. These pathological changes simultaneously involve multiple adjacent lobules. HRCT images appear as single or multiple ground glass opacities locating peripheral lung elds accordingly. With the progress of disease, multiple patchy ground glass opacities increase and fuse to round or aky lesions without the distribution of the pulmonary segments. Lung involvements with a peripheral predominance of ground glass opacities are also the primary CT ndings of SARS21 and MERS22. On CT images, pleural effusions could be found in few patients with SARS or MERS, but lymphadenopathy could not be seen in any patient23,24. The absence of pleural effusions in most patients was also characteristic of COVID-19 in our study. According to previous research, early onset of pleural effusion with a higher pulmonary CT scores was a sign of poor prognosis25. Focal ground glass opacities and consolidations located in peripheral subpleural elds rapidly progressed to almost the whole lung and were responsible for ensuing acute respiratory distress syndrome (ARDS)25. ARDS commonly occurred in SARS and MERS26. Therefore, COVID-19 patient with pleural effusion and a short incubation period should be given adequate attention and aggressive treatment to prevent rapid progression even to ARDS. 2019-nCoVs distribute over respiratory mucosa, infect other cells, cause a cytokine storm in the body, produce a chain of immune responses, and generate changes in immune cells and peripheral white blood cells. Noticeably, 2019-nCoV mainly attacks lymphocyte, particularly T lymphocyte, thus, a decrease in lymphocyte count is a common laboratory test nding of COVID-19 infection. Degree of lymphocytopenia might be a critical predictive factor associated with disease severity and mortality7. The present study demonstrated the negative correlation between the number of lobes involved or the CT lung severity score and lymphocyte count. The CT lung severity score and the number of lobes involved may be the surrogate biomarkers in predicting disease severity of COVID-19.
There were several limitations in our study. First, the sample size of the present study was relatively small. Only 3 cases had followed-up CT scans. The progressions and outcomes of COVID-19 have not been accurately assessed. Second, our cases were all adults. The CT ndings of COVID-19 in children have not been evaluated. Third, because not all of cases were from Wuhan, our results may be incomprehensive.
In conclusion, the pulmonary CT of COVID-19 predominantly demonstrates multiple ground glass opacities, with a preference for bilateral lungs involvement in peripheral regions. Pleural effusion and mediastinal lymphadenopathy are rare imaging manifestations in COVID-19. Identifying these imaging characteristics in patients living in or having travelled to Wuhan or other epidemic areas can be helpful for early and timely diagnosis of COVID-19. The CT lung severity score and the number of lobes involved have the negative correlation with lymphocyte count. Thin-section HRCT may play an important role in predicting disease severity of COVID-19.

Materials And Methods
This retrospective study was granted by the Institutional Ethical Review Board of our hospitals with a waiver of informed consent.
Clinical data and CT Protocol 32 patients with con rmed COVID-19 presented to two hospitals in two cities underwent plain chest CT during January 17 to February 6, 2020. Nine patients were imaged with 1.25-mm slice thickness on a GE optima 540 CT scanner (GE Medical Systems, Milwaukee, Wis). Twenty-three patients were imaged with 0.625-mm slice thickness on a GE optima 520 Pro CT scanner (GE Medical Systems, Milwaukee, Wis) and a Siemens SOMATOM CT scanner (Siemens Healthineers, Erlangen, Germany). All patients were supine, head advanced, and end-inspiration breathhold during the examination. Clinical data were collected including age, sex, symptoms, travel and exposure history. Peripheral blood routine test data were also collected. Notably, case selection of this study was consecutive in the two hospitals separately.

CT images analysis
All chest CT images were gathered and reviewed by two radiologists with more than 8 years' experience in thoracic imaging diagnosis independently. Final decisions were reached by consensus. When the result was inconsistent, the nal decision was made by a chief radiologist with 22 years' experience in imaging diagnosis.
The CT features of all the 32 patients were evaluated as follows: ( ) Pulmonary lesions: The location, morphology, number and density of the lesions were analyzed. Moreover, the number of lobes involved by lesions was also analyzed. ( ) Accompanying signs: The presence of pleural effusion, mediastinal lymphadenopathy (the lymph node in size of ≥10 mm in short-axis dimension). ( ) The CT lung severity score: Based on tracheal carina and inferior pulmonary veins, each lung was divided into three zones including upper middle, and lower zones. For each lung zone, the severity score ranged from 0 (normal) to 4. The 0-4 represented normal (0%), minimal (1%-25%), mild (26%-50%), moderate (51%-75%), or severe (76%-100%) respectively. The sum of scores afforded total lung involvement (maximal CT score was 24 for both lungs). ( ) Progression of disease: 3 patients underwent a follow-up chest CT during our study time window. These images were assessed to the change and progression over time.

Statistical analysis
Statistical analysis was performed using SPSS software for Windows (version 17.0, SPSS Inc., Chicago, IL). The Pearson correlation test was executed to assess the correlation between the blood test data and CT data. The tests were two-tailed, and P-values <0.05 were considered to indicate statistically signi cant difference.
Data Availability Statement: The authors declare data of this manuscript is authentic and reliable, and can provide a URL or other unique identi er in the manuscript upon any requirement from the reader.   Table 4. Correlation between CT Data and Laboratory Data. Note: All data were analyzed by using the Pearson correlation test. Figure 1 Image in a 47-year-old male with unknown exposure history of COVID-19 patients, presenting with fever and cough. Axial non-contrast HRCT CT image shows multiple bilateral ground-glass opacities with consolidation with a peripheral distribution (red arrows).

Figure 2
Image in a 43-year-old male having travelled to Wuhan, presenting with fever and headache. Axial noncontrast HRCT CT image shows vascular bundles congestion (black arrow) and irregular intralobular lines (red arrows) within a round ground-glass opacity. Bilateral pleural effusion is seen (white arrows). (c) Axial non-contrast HRCT image (1/21/2020) shows enlarged and increased ground glass opacities accompany with more solid areas (red arrows) in bilateral lungs, hinting disease progression.

Figure 4
Scatter plots revealing correlation between CT data and lymphocyte count. (a) There is a negative correlation between the number of lobes involved and the lymphocyte count(r=-0.367, P=0.039). (b) This