Corona virus disease 2019: hysteresis effect of chest CT and the correlation of its severity with lymphocyte count in peripheral blood CURRENT

Background Chest computed tomography (CT) has been accepted to provide reference for the diagnose and assessment the severity of Corona Virus Disease 2019 (COVID-19). Decrease in the counts of lymphocyte and leukocyte is used as the diagnostic indicator of suspected COVID-19 cases. However, there is few study on exploring the hysteresis of chest CT changes and the predictive role of lymphocyte count in peripheral blood before treatment in the severity of the disease. Methods A retrospective analysis was carried out focusing on the data of patients tested to be positive for RNA nucleic acid test of SARS-CoV-2 with nasopharyngeal swabs in 4 hospitals. An independent assessment was performed by one clinician using the DEXIN FACT Workstation Analysis System, and the assessment results were reviewed by another clinician. Furthermore, the mean hysteresis time was calculated according to the median time from progression to the most serious situation to improvement of chest CT in patients after fever relief. The optimal scaling regression analysis was performed by including variables with statistical significance in univariate analysis. In addition, a multivariate regression model was established to investigate the relationship of the percentage of lesion/total lung volume with lymphocyte and other variables. Results In the included 166 patients with COVID-19, the average value of the most serious percentage of lesion/total lung volume was 6.62, of which 90 patients with fever had an average hysteresis time of 4.5 days after symptom relief, with a similar trend observed in those without fever. Multivariate analysis revealed that lymphocyte count in peripheral blood and transcutaneous oxygen saturation decreased with the increase of the percentage of lesion/total lung volume. Meanwhile, age, fever and C-reactive protein exhibited no such effect in the established model. Conclusions There is a hysteresis effect in the improvement of chest CT image in relative to fever relief in patients with COVID-19. Besides, the percentage of lesion/total lung volume of chest CT correlates negatively with lymphocyte count in peripheral blood and transcutaneous oxygen saturation. Findings in our study may contribute to understanding the disease status of patients with COVID-19 and grasping the opportunity of treatment by clinicians. Kruskal-Wallis H test (categorical variables) and Spearman correlation analysis (measurement variables). The presence of collinearity was determined if the correlation


Abstract
Background Chest computed tomography (CT) has been accepted to provide reference for the diagnose and assessment the severity of Corona Virus Disease 2019 . Decrease in the counts of lymphocyte and leukocyte is used as the diagnostic indicator of suspected COVID-19 cases. However, there is few study on exploring the hysteresis of chest CT changes and the predictive role of lymphocyte count in peripheral blood before treatment in the severity of the disease.
Methods A retrospective analysis was carried out focusing on the data of patients tested to be positive for RNA nucleic acid test of SARS-CoV-2 with nasopharyngeal swabs in 4 hospitals. An independent assessment was performed by one clinician using the DEXIN FACT Workstation Analysis System, and the assessment results were reviewed by another clinician. Furthermore, the mean hysteresis time was calculated according to the median time from progression to the most serious situation to improvement of chest CT in patients after fever relief. The optimal scaling regression analysis was performed by including variables with statistical signi cance in univariate analysis. In addition, a multivariate regression model was established to investigate the relationship of the percentage of lesion/total lung volume with lymphocyte and other variables.

Results
In the included 166 patients with COVID-19, the average value of the most serious percentage of lesion/total lung volume was 6.62, of which 90 patients with fever had an average hysteresis time of 4.5 days after symptom relief, with a similar trend observed in those without fever. Multivariate analysis revealed that lymphocyte count in peripheral blood and transcutaneous oxygen saturation decreased with the increase of the percentage of lesion/total lung volume. Meanwhile, age, fever and C-reactive protein exhibited no such effect in the established model.

Conclusions
There is a hysteresis effect in the improvement of chest CT image in relative to fever relief in patients with COVID-19. Besides, the percentage of lesion/total lung volume of chest CT correlates negatively with lymphocyte count in peripheral blood and transcutaneous oxygen saturation. Findings in our study may contribute to understanding the disease status of patients with COVID-19 and grasping the opportunity of treatment by clinicians.

Background
Corona Virus Disease 2019 (COVID-19) is featured by strong infectivity, rapid development, and high incidence in Spring and Autumn, showing an epidemic trend in the world recently. With a high de nition, computed tomography (CT) facilitates the display of ground glass-like shadows and foci in the overlapped area behind the heart for COVID-19 patients [1]. Despite limitations in qualitative diagnosis and quanti cation of lesions, CT is still the most widely used tool to diagnose and assess diseases worldwide. The weight of CT reached 41.67% (5/12) when it was used to evaluate the severity of COVID-19 [2,3,4].
In general, imaging severity of diseases can be graded by estimating the range of "abnormal" lung eld.
In the past, Seaton et al. [5,6] described an anatomical site-based classi cation, besides, Ralph et al. [7] and Pan et al. [8] in the later stage proposed an evaluation method of using integral. These methods, however, are not only lack of accuracy, but also mainly used in chronic infectious diseases. With the development of medical imaging, quantitative CT analysis technology has been used in clinic in recent decades [9], mainly for the evaluation of pulmonary nodules. It will be totally a new perspective to apply CT to COVID-19, an acute infectious disease. It still requires further exploration on how to recognize the correlation between the degree of quantitative computerized image changes and clinical indicators.
Simultaneously, it was found in our clinical practice that the lesion of chest CT was still progressing in some patients with fever relief. When would the relative hysteresis period of imaging end? It dose confuse our understanding of the disease clinically.
With respect to the above, the purpose of this study is to determine the hysteresis time of CT images in subjects with fever and compare the change trend of quantitative CT parameters with that of subjects without fever; and to clarify relevant factors affecting the percentage of lesion/total lung volume in patients with COVID-19.

Study sites and patients
The objects of study were 170 patients from 4 hospitals in 2 provinces of China from January 2020 to March 2020. According to the COVID-19 Diagnosis and Treatment Plan issued by the National Health Commission, patients were provided with antiviral treatment of interferon-α combined with Lopinavir/Ritonavir or Arbidol, chest CT scan every 3-5 days, sampling of nasopharynx swabs and routine blood examination (routine blood test, liver function, etc.). All enrolled patients were detected to have positive nucleic acid test results for severe acute respiratory syndrome coronavirus (SARS-CoV)-2 by real-time uorescent RT-PCR of nasopharyngeal swabs. During the treatment period, patients were excluded for those < 18 years old, with positive HIV antibody, and dead within one week after diagnosis.
Percentage of lesion/total lung volume Chest CT images were scanned by an imaging technician according to the uni ed parameters (see Additional le 1). The imaging technician was responsible for the immediate assessment of the "readability" of the image and re-scanning of the unquali ed image. The original data of collected images was then imported into the "Digital Lung" Pneumonia Quantitative Software (DEXIN FACT Workstation Analysis System) in a format of Digital Imaging and Communications in Medicine (DICOM), of which the software has been veri ed [10]. Pulmonary lesions of COVID-19 patients were automatically segmented, extracted and quanti ed. Subsequently, anther manual identi cation (especially ground glass-like shadows), segmentation and deletion of the lesion was carried out according to the preset inclusion and exclusion criteria (Table 1). Another doctor was involved to review the results carefully. Besides, a third senior doctor blinded with the assessment was invited to make the nal determination in the case of inconsistent lesion judgment. The volume of the whole lung (ml) and that of the lesion after manual correction (ml) were automatically calculated by the system to obtain the percentage of the lesion/total lung volume. Finally, following repeated chest CT examination, the maximum value of the percentage of lesion/total lung volume was selected for univariate and multivariate analysis.  Chest CT hysteresis time: on the basis of calculation from the rst day of admission, it was de ned as the average duration (days) of fever by subtracting the average duration of fever after admission from the median time of lesion/total lung volume percentage reduction ≥ 10% compared with the previous time in subjects with fever (nursing record of body temperature ≥ 37.3℃). Cough and asthma symptoms were not included in this analysis since the two symptoms had no clear time line for relief, which might hence affect the accurate quantitative results. Data related to routine blood test, C-reactive protein (CRP), albumin and transcutaneous oxygen saturation (without oxygen inhalation) subjected to the rst test results of patients after admission. As part of the clinical data, the inpatients were tested for HIV, in combination of collecting medical history of diabetes, hypertension, etc.

Statistical analysis
Multivariate analysis of variance was used to compare the percentage of lesion/total lung volume in subjects with and those without fever at different times. Univariate analysis for baseline characteristics (age, gender, body mass index, primary/secondary cases, leukocytes, lymphocytes and presence of underlying diseases) and the percentage of lesion/total lung volume were analyzed by Mann-Whitney U test, Kruskal-Wallis H test (categorical variables) and Spearman correlation analysis (measurement variables). The presence of collinearity was determined if the correlation coe cient of the two variables was > 0.8 or the tolerance was < 0.1 before data conversion. Series mean was used to replace the missing value. Finally, the variables with statistical signi cance in univariate analysis (P < 0.05) were involved for optimal scaling regression analysis. SPSS 22.0 (IBM Corp., Armonk, NY, USA) and GraphPad Prism 8.0.2 (GraphPad Software, La Jolla, CA, USA) were used for statistical analysis in this study.

Results
Out of 170 con rmed patients with COVID-19 in 4 hospitals, 4 cases were excluded, and 166 were nally included in the study (Fig. 1 Changes of DICOM analysis parameters at different times Table 2 shows the changes of parameters such as total lung volume and lesion volume at different times, in which the percentage of lesion/total lung volume was the largest within 3-6 days after admission and tended to improve in 7-10 days. Figure 2 displays the change of representative case. Chest CT showed that the median percentage of lesion/total lung volume was 6.62, with the interquartile range of between 3.59-10.25. CT computed tomography, HU Houns eld unit, IQR interquartile range. a The change range of the percentage of lesion/total lung volume compared with the previous time. "↑" means increase and "↓" means decrease Image hysteresis time of patients with fever The median disappearance time of fever was 3.5 (2.00-5.25) days in the 90 subjects with fever after admission. The rst ≥ 10% decrease in the percentage of lesion/total lung volume compared with the previous time was in 7-10 days, with a median time of 8 days. Hence, the image hysteresis time of subjects with fever was measured to be 8.0-3.5 = 4.5 days. Analysis of variance indicated that there was no difference in the change of the percentage of lesion/total lung volume in subjects with and those without fever and all subjects at different times (F = 1.839, P = 0.161) (Fig. 3). Besides, there was obvious difference in the percentage of lesion/total lung volume at different times (all P < 0.001), without the discovery of interaction between time and group (see Additional le 2).
In uential factor of the percentage of lesion/total lung volume According to the results of univariate analysis, age, fever, absolute value of lymphocyte, CRP and transcutaneous oxygen saturation were signi cantly related to the percentage of lesion/total lung volume (all P < 0.05) (Tables 3 and 4). These 5 variables were thus included in the multivariate analysis. Corresponding correlation coe cients were < 0.8 (age and lymphocyte count were the largest dependent variable with a coe cient of -0.348), and there was no variable with the tolerance < 0.1. The results showed that the adjusted R 2 was 0.250, indicating the existence of statistical signi cance of the established regression model (F = 10.675, P < 0.001). Meanwhile, lymphocyte count and transcutaneous oxygen saturation were independent risk factors for the percentage of lesion/total lung volume (all P < 0.05) (

Discussion
In our study, based on "Digital Lung" Pneumonia Automatic Detection/Quantitative System, quantitative detection indexes of COVID-19 patients were obtained by computer-aided detection and manual correction, including total lung volume, lesion volume, and average density of lesions. Among them, lesion volume is the most basic feature of lesions, which directly re ects the actual size of the lesion, total lung volume is the basic general description of whole lung in the patient, and density can help to judge the composition and nature of lesions. The percentage of lesion/total lung volume re ects the severity of the disease in part and in general.
According to the results of this study, 74.10% (123/166) of patients with COVID-19 reported fever, with the median time of fever of 5.0 (4.0-8.0) days before admission. After admission, 54.22% of the patients were con rmed to have fever, and the median time of remission was 3.5 (2.0-5.3) days. Furthermore, images of chest CT began to improve with an average hysteresis of 4.5 days after fever relief. It was 13 days when it was calculated since the onset of fever, which was similar to those reported by Jin et al [12] and Pan et al [8] that the radiological manifestation of chest CT entered the dissipation stage at 2-3 weeks. Simultaneously, ndings in our study suggest that: (1) clinical symptoms are still important clues for early assessment of the disease; and (2) the possibility of developing to severe lesions should be alerted if there is still progression in chest CT of patients beyond this median time.
Patients with fever had a higher percentage of lesion/total lung volume. It is speculated that fever attenuates the immune response of cytokines by inhibiting the expression of tumor necrosis factor-α at ribonucleic acid level [13], thus inhibiting the immune response of natural killer (NK) cells [14] and cytotoxic T lymphocytes (CTL) [15], resulting in a higher risk of lesion spread. The present study further analyzed the relationship between fever and lymphocyte, and it was found that lymphocyte count was lower in patients with fever (Fig. 6).
In the subsequent univariate analysis, both age and CRP level showed positive correlation with the percentage of lesion/total lung volume, which was consistent with the higher intensive care unit admission rate [16] and more pulmonary consolidation [17] reported by Wang et al. Elevated CRP in patients with COVID-19 was revealed to be associated with acute lung injury [18,19]. However, age and CRP had little effect in this study, which were not statistically signi cant in the nal multivariate analysis.
In addition, no signi cant difference was discovered in the impact of travel or residence history of Hubei Province on the area of lesion, which was consistent with the domestic report [20]. Meanwhile, 19.28% (32/166) patients had underlying diseases, which was lower than that reported by Huang et al. [21]. It is possible that the elderly developed the disease in a pattern of family-centered rather than communitycentered onset in other regions outside Wuhan. It is thus not related to the elderly patients with more underlying diseases.
Peripheral blood lymphocytes include T, B, NK, CTL and other cells. NK cells play a crucial role in the rst phase of immune response against infections, which is quite important in controlling the viral load of lung in the early stage of virus infection [22,23,24]. Studies have demonstrated that the severity of pulmonary lesions is positively correlated with the viral load in the body [19], while the latter one exhibits an intimate correlation with the decline of immune cells [25]. Therefore, the lymphocyte count re ects the degree of pulmonary lesions to a certain extent. In the sequential studies, it was found that the activation of CD4 and CD8 T cells was impaired in SARS-CoV infected patients [26,27], and the decrease of T lymphocyte count was strongly correlated with the severity of lesion in the acute phase [28,29]. In our study, counts of lymphocyte in 37.95% (63/166) patients were lower than the lower limit of normal value (1.1 × 10 9 /L), which was close to the lymphocyte decline rate reported by Assiri et al. [30] in Middle East respiratory syndrome coronavirus (MERS-CoV). It suggests that lymphocyte count presents similar change in COVID-19 compared with that in SARS-CoV and MERS-CoV, which is an independent negative risk factor for the percentage of lesion/total lung volume. It may provide a strong scienti c basis for early identi cation of severe patients.
Furthermore, oxygen saturation re ects the percentage of oxyhemoglobin in total hemoglobin. Its decrease suggests that the alveolar injury and the concomitant increase of cellular bromyxoid exudates may further lead to the ventilatory disorder of patients [31], and the degree of decrease may indirectly re ect the degree of pathological changes. In this study, 16.27% (27/166) patients showed the oxygen saturation of < 95%, which was different from 9.40% reported by Yang et al. [20]. It may be explained by the disunity of detection methods and oxygen inhalation.
The study still has certain limitations as described below.
(1) A descriptive study was carried out only for the image hysteresis of chest CT, without further time series analysis on changes of body temperature and the percentage of lesion/total lung volume. (2) In this study, the area of lesion was the main criterion for lesion identi cation. Despite the presence of statistical signi cance regarding changes of CT value of lesions at different times (Friedman test, χ 2 = 179.412, P < 0.001), CT value is not constant and can be changed with power supply, scanning parameters, temperature and adjacent tissues. Therefore, it is necessary to combine the actual situation for judgment in the application. Availability of data and materials The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Authors' contributions
KL and Z-CJ designed the study, performed data analysis and data interpretation, and wrote the manuscript. WW, C-WZ and LW analyzed the chest CT. Y-NZ, R-YR and TL were instrumental in supervising and collecting the images. All authors have read and agreed with the nal manuscript.

Ethics approval and consent to participate
This study was approved by the Ethics Committee of the Ankang Central Hospital (ECACH-2020002). The patient's informed written consent to analysis of their medical records was waived due to the retrospective nature of this study. No further permission from the hospital was required.

Consent for publication
Not applicable

Competing interests
The authors declare that they have no competing interests.
Publisher's Note Changes in lesions at different times in COVID-19. A male patient aged 37 years old had low fever (maximum body temperature of 37.8℃) and dry cough for 11 days, with the initial diagnosis of type 2 diabetes. Chest CT examination: scattered occulent, patchy and slightly high-density shadows with unclear boundary in both lungs. Red area represents the lesion volume, Green and blue represent the normal lung volumes of the right and left lungs, respectively. a, b, c, d and e were the images on the 2nd, 6th, 10th, 15th and 21st day after treatment, respectively Scatter plot of the percentage of lesion/total lung volume (x-axis) and transcutaneous nger pulse oxygen saturation (y-axis). Linear correlation analysis showed a decrease trend of transcutaneous nger pulse oxygen saturation with the increase of percentage ratio