Chest CT imaging features of critically ill COVID-19 patients CURRENT STATUS: POSTED

Objectives To analyze the findings of computed tomography (CT) imaging in critically ill patients diagnosed with coronavirus disease 2019 (COVID-19). Methods This retrospective study reviewed 60 critically ill patients (43 males and 17 females, mean age 64.4±11.0 years) with COVID-19 pneumonia who were admitted to two different clinical centers. Their clinical and medical records were analyzed, and the chest CT images were assessed to determine the involvement of lobes and the distribution of lesions in the lungs between the patients who recovered from the illness and those who died. Results Patients were significantly older in the death group (10/60, 16.67%) than in the recovery group (50/60, 83.33%) (p=0.044). C-reactive protein (CRP) (67.9±50.5 mg/L) was significantly elevated in the death group as opposed to the recovery group (p<0.001). The neutrophil-to-lymphocyte ratio (NLR) was higher in the death group when compared with the recovery group (p=0.030). Involvement of five lung lobes was found in 98% of the patients, with medial or parahilar area involvement observed in all the death patients. Ground-glass opacities (97%), crazy-paving pattern (92%) and air bronchogram (93%) were the most common radiological findings. Presence of emphysema was more prevalent in the death group than in the recovery group (30% vs 2%, p=0.011). Conclusions The degree of lung involvement and lesion distribution with dominance in the medial and parahilar pulmonary areas were more severe in the death patients than in those who recovered. Patient’s age, emphysema, CRP and NLR could be combined with CT to predict the disease outcomes. obstructive pulmonary disease, CRP: C-reactive protein, WBC-white blood cell, LYM-lymphocyte, NEUT-neutrophils, NLR-neutrophil-to-lymphocyte ratio. Ten patients died in this study cohort, establishing the mortality of critically ill COVID-19 pneumonia patients to be 16.67%. The condition of the remaining 50 patients improved, and they were discharged from the hospitals. No significant difference was found in gender, time course of symptoms prior to hospitalization, and comorbidity between the recovery and death groups.


Introduction
From early December 2019, coronavirus disease 2019 (COVID- 19), which is caused by the novel coronavirus (2019-nCoV), has rapidly spread from Wuhan to other regions of China and countries around the world. According to the World Health Organization (WHO) report, by 2 March 2020, there were 88948 confirmed cases globally, including 79968 in China, and 3043 deaths worldwide [1].
Management of critically ill patients is important to reduce the mortality of COVID- 19. In China, the reported incidence of critical illness in COVID-19 patients was 17.7% in Wuhan, 10.4% in the Hubei province, and 7.0% in areas outside the Hubei province [2]. These figures necessitate attention since the incidence of critical illness among the Chinese medical staff afflicted with COVID-19 was 14.6% [2]. According to a recent study by Guan et al. who reported the clinical characteristics of  in China through an analysis of 1099 patients, 173 (15.7%) had severe disease with a mortality of 8.1%, which was significantly higher than that in the non-severe patients (0.1%) [3]. The mortality of critically ill COVID-19 patients in China was between 38.5% and 49.0% [2,4,5]. Thus, it is imperative to recognize both the clinical and imaging characteristics, thus achieving superior patient management.
Recently, several studies have reported chest computed tomography (CT) imaging features and changes during recovery in COVID-19 pneumonia patients without acute respiratory distress syndrome [6][7][8][9][10][11][12][13]. The common conclusion arising from these investigations is that CT is a useful imaging modality in the diagnostic evaluation of abnormal lung changes in the patients [11,12,14,15]. However, CT manifestations in critically ill patients have not been described in literature. Hence, the purpose of this study was to analyze the chest CT findings in a group of critically ill COVID-19 pneumonia patients with the aim of determining the clinical and imaging features that can be used to predict the future disease outcome.

Materials And Methods Patients
We retrospectively reviewed the medical records of 60 critically ill COVID-19 pneumonia patients who were admitted to a hospital in Wuhan, Hubei province, and another in Huaihua, Hunan province, between 9 January 2020 and 19 February 2020. Non-contrast chest CT examinations were performed in all the patients. They were diagnosed by the local Centre for Disease Control and Prevention (CDC) (Hubei and Hunan Provinces) by using the real-time reverse-transcription-polymerase-chain-reaction (RT-PCR) assay with samples of bronchoalveolar lavage, endotracheal aspirate, nasopharyngeal swab, or oropharyngeal swab. According to the criteria for clinical severity of confirmed COVID-19 pneumonia (Table 1) [16], the patients of severe and critical type were defined as being critically ill.
All CT scans were performed within 24 h for patients who met the clinical severity criteria. No exclusion criteria were applied since the selection criteria were strictly followed to include only the severely ill patients.

Chest CT Evaluation
The CT images were independently reviewed by two cardiothoracic radiologists (having 20 years and The terms of abnormal imaging appearances were defined according to the guidelines provided by the Fleischner Society [17]. GGO opacity was defined as hazy increased lung density, with indistinct margins of bronchus and pulmonary vessels. Consolidation was defined as increased pulmonary parenchymal attenuation, with the margins of the bronchus and the pulmonary vessels being obscured. Crazy-paving pattern was defined as GGO opacity combined with reticulation or/and interlobular septal thickening. An air bronchogram is an air-filled bronchus seen clearly with low density on a background of GGO or consolidation opacity. The margin of the lesion was defined as sharp or blurred. Thoracic lymphadenopathy was defined as the short-axis dimension of lymph node ≥ 10 mm.

Statistical Analysis
The data were analyzed using SPSS 25.0 (SPSS, Inc., Chicago, IL, USA). The continuous variables were expressed as mean ± standard deviation (SD). The comparison of discrete variables between the recovered and death groups was done by using X 2 test with continuity correction. The continuous variables were compared using Mann-Whitney U test. The differences were considered statistically significant at p ≤ 0.05.

Patient Characteristics
Of the 60 patients, 43 were males (71.7%), and their mean age was 64.4±11.0 years. Patients in the death group were older than those in the recovery group by an average of 7 years. The mean time between symptoms onset to admission was 8.9±5.0 days ( Table 2). Hypertension (37%) was the most common comorbidity in these critically ill patients, followed by coronary heart disease (23%). Fever (83%) and cough (77%) were the most common symptoms, followed by phlegm (55%), fatigue (55%) and loss of appetite (47%). C-reactive protein (CRP) (67.9±50.5 mg/L) was elevated in all the patients, but it was significantly greater in the death group than in the recovery group (p<0.001). Similarly, the NLR ratio was significantly higher in the death group when compared with the recovery group (p=0.03). COPD-chronic obstructive pulmonary disease, CRP: C-reactive protein, WBC-white blood cell, LYMlymphocyte, NEUT-neutrophils, NLR-neutrophil-to-lymphocyte ratio.
Ten patients died in this study cohort, establishing the mortality of critically ill COVID-19 pneumonia patients to be 16.67%. The condition of the remaining 50 patients improved, and they were discharged from the hospitals. No significant difference was found in gender, time course of symptoms prior to hospitalization, and comorbidity between the recovery and death groups.

CT Findings
Bilateral lungs were involved in all the patients. Almost all the 60 patients (98%) exhibited involvement of all 5 lobes, while only 4 lobes were involved in 1 patient. In the death group, the lesions included the peripheral, intermediate, and medial areas for all the patients. Medial or parahilar area involvement was highly prevalent in the death group (recovery patients: 27/50, 54%; death patients: 10/10, 100%, p=0.018) (Fig 1). The degree of lung involvement was more severe in the death group (recovery patients: 2.0±0.7; death patients: 3.3±0.5, p<0.001), with no significant difference in the consolidative degree of the lesions (recovery patients: 1.6±1.1; death patients: 1.5±1.4, p=0.662) ( Table 3). GGO appearances were noted in 97% of the patients (Fig. 2), with consolidation in 68% and linear opacities in 22% (Fig 3). There was no significant difference between the recovery and death patients in the lesion types. Crazy-paving pattern (92%) and air bronchogram (93%) were discernable in almost all the patients ( Fig. 4 and 5). The margin of the lesions was blurred in 98% of the patients and clear in only 20% (Fig. 6). No significant difference was noticed between the recovery and death patients in the presence of the above-mentioned signs.
The presence of emphysema was more common in the death group (recovery patients: 2%; death patients: 30%, p=0.011). Only 1 (2%) patient exhibited fibrosis in the CT images. Calcified lesions were found in 22% of the patients. Pleural effusion was evident in 25% of the patients and pericardial effusion in 3%. Lymphadenopathy was not found in any of the patients.

Discussion
In this study, we focused on the analysis of basic clinical information, comorbidity, symptoms, laboratory results, clinical outcomes and CT findings in 60 critically ill COVID-19 pneumonia patients.
This investigation is different from the previous ones which had included different categories of COVID-19 patients and compared mild with moderate or severe, severe with non-severe, and intensive care unit (ICU) with non-ICU patients [2,[18][19][20][21]. Of the 60 patients in our study, 10 (16.67%) died; thus, the mortality rate is similar to the 15% that was reported by Huang et al. [4] but much higher than the 1.4% for all the patients and 8.1% for the severe patients as documented by Guan et al. [3]. However, most of the patients (93.6%) remained in the hospital in the study by Guan et al.; therefore, the clinical outcomes were unknown at the time of publication despite the analysis of a large sample size. The mortality rate in our cohort is lower than the 38.5 and 49% that was reported in China [2,4,5,22]. The low mortality recorded in these investigations could have been due to enrolment with the exclusion of critically ill COVID-19 patients. It could also have been due to the fact that not all the patients received CT scans within 24 h, considering the emergency of the situation.
Since 11.1-19%; 5 lobes involvement: 38-44.4%) [6,7]. However, in this study of the critically ill patients, involvement of all 5 lobes was observed in 98% of the patients. A larger proportion of intermediate (87%) and medial (62%) areas was involved in our recruited critically ill patients, which was in contrast with the predominant peripheral involvement in the mild and moderate patients [7,11,13,24]. Among the critically ill patients in this study, the mean degree of lung involvement was 2.2±0.9 with a range of 0-4, which represents nearly 50% of the lung field being involved. In the death group, the mean score was 3.3±0.5, indicating a higher degree of lung involvement. In the research by Chung et al. [7], a total of the lung severity scores of mild and moderate patients were calculated and a summation of each lobe score (with similar 0-4 scales) was performed to determine the degree of involvement of the lung field. Their results showed that the mean total lung severity score was 9.9 in a range of 0-20. Our results aid in predicting the extent of disease in these critically ill patients by analyzing the degree of lung involvement based on chest CT images.
According to the MuLBSTA scoring system [25], multilobular infiltrates, lymphocyte ≤ 0.8×10 9 /L, bacterial coinfection, acute-smoker, quit-smoker, hypertension and age ≥60 years are the mortality risks for viral pneumonia. In this work, although 5 lobes were infiltrated in most patients, medial or parahilar area involvement and the degree of lung involvement were significantly higher in the death group. Involvement range and degree might be the potential risk predictors on CT images in the COVID-19 pneumonia patients. Our patient sample is different from those in other studies as we evaluated the clinical and imaging features in the critically ill patients with pneumonia. Since emphysema was more common in the death group, it could be hypothesized that an underlying lung disease may also affect the clinical outcome. Furthermore, patients with low lymphocyte count, hypertension and old age appeared more frequently in the death group.
The NLR was identified as the independent risk factor for predicting critical illness in the COVID-19 pneumonia patients, with 3.13 serving as a good cut-off value [26]. In this research, the average NLR of the 60 patients was 9.7±9.5, which is significantly higher than the recommended value of 3.13.
Moreover, the NLR was significantly higher in the death patients (nearly double the value of the recovery group), indicating its potential to predict not only critical illness but also death in the severely ill patients. reported that some recovered patients may still be carriers of the virus [29]. Therefore, long-term follow-up of these patients is necessary, which should be the focus of future research. Lastly, no autopsy was performed in the death patient group. Upon comparison with the pathological results, additional and a more precise interpretation of the CT image signs will be available in the future work.
In conclusion, this study involving critically ill COVID-19 patients has revealed that ground-glass opacities, crazy-paving pattern and air bronchogram represent the most common findings, with more of the pulmonary lobes involved in the patients. Medial and intermediate area involvements in the lungs were more often seen in the death group than in the recovery group. Additionally, some clinical and laboratory factors such as patient age, co-existing emphysema, CRP and NLR could be used to predict the disease outcomes as they were significantly higher in the patients who succumbed to the disease than in those who recovered from it.

Funding information
The authors declare that they have not received any funding related to this work. Peripheral, intermediate, and medial areas were seen to be simultaneously involved on CT, with lung involvement and lesion consolidation degrees of 51%-75%.