Temporal Changes of CT Findings in 90 Patients with COVID-19 Pneumonia: A Longitudinal Study

Introduction

The outbreak of coronavirus disease 2019 (COVID-19) originated in Wuhan, China, began in December 2019, and has been continuing for more than 3 months (1). The disease quickly spread across China and beyond, and, as of March 8, 2020, a total of 105 586 confirmed cases, including 3584 deaths, have been reported worldwide (2).

CT plays a central role in the diagnosis and management of COVID-19 pneumonia (3). Previous efforts to analyze CT manifestations have continued. Some studies demonstrated the CT findings to be diverse, with the main abnormalities including ground-glass opacity and consolidation (4–6). Pan et al (7) described the evolution of CT findings in 21 patients with mild COVID-19 pneumonia. A better understanding of the progression of CT findings in COVID-19 pneumonia may help facilitate accurate diagnosis and disease staging. Thus, we performed a longitudinal study to analyze the serial CT findings in patients with COVID-19 pneumonia for temporal change.

Materials and Methods

This prospective observational study was approved by the institutional review board of Union Hospital, Tongji Medical College, Huazhong University of Science and Technology. The requirement to obtain informed consent was waived by the review board due to the observational nature of the study and the epidemic of COVID-19 as an emergency public health event.

Patients

Patients who were admitted to the isolation wards of Union Hospital with suspected COVID-19 pneumonia from January 16 to February 17, 2020, were screened by using the following inclusion and exclusion criteria.

The inclusion criteria were as follows: (a) at least one positive real-time reverse-transcriptase polymerase chain reaction (RT-PCR) result for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) was obtained with oropharyngeal swabs before or after admission, (b) at least one CT scan showed lung abnormalities before or after admission, and (c) electronic records were available. No specific exclusion criteria were applied.

Treatment and Discharge Criteria

Treatment and discharge criteria for patients with COVID-19 pneumonia were in line with the Guidelines for the Diagnosis and Treatment of COVID-19 Pneumonia published by the National Health Commission of the People’s Republic of China (3). The discharge criteria were as follows: (a) patients were afebrile for at least 72 hours, (b) the respiratory symptoms showed significant improvement, (c) there was evidence of improvement on chest CT scans or radiographs, and (d) there were two consecutive negative SARS-CoV-2 RT-PCR results at least 24 hours apart.

After admission, CT examinations were at the discretion of the treating clinicians as appropriate for the clinical scenario. We followed up the patients until they were discharged from the hospital, until death, or until the end of the study (March 8, 2020) if they were still in the hospital.

CT Protocol

Unenhanced chest CT scans were obtained with the patient in the supine position, and scanning was performed at end-inspiration. To minimize motion artifacts, patients were instructed on breath-holding. CT examinations were performed by using multidetector CT scanners (Discovery 750HD [GE Medical Systems, Milwaukee, Wis] or Activion 16 [Toshiba, Tokyo, Japan]). The scans were acquired by using 120 kVp and adaptive tube current, and the images were reconstructed as axial images with 1.25-mm section thickness and 1.25-mm intervals. However, due to the large number of patients presenting for emergency CT and quick review, 5-mm section thickness and 5-mm gaps were sometimes used for reconstruction with the Activion 16 unit for a faster transfer of images to the picture archiving and communications system. Patients were assigned to whichever scanner that was available at the time of imaging.

The CT scans obtained before admission for each patient were also used for analysis.

CT Image Interpretation

Five experienced radiologists in two groups read the CT images (group 1: Y.W., C.L., and X.Z., with 11, 12, and 3 years of experience in radiology, respectively; group 2: C.D. and Q.R., with 5 and 6 years of experience in radiology, respectively). Multiple CT scans obtained in a single patient were reviewed by the same group, and the decision was reached in consensus.

The CT patterns were described according to the terms defined by the Fleischner Society and peer-reviewed literature on viral pneumonia (8,9). The CT images were assessed for the presence of ground-glass opacity (hazy areas of increased attenuation without obscuration of the underlying vasculature), consolidation (homogeneous opacification with obscuration of the underlying vasculature), reticular pattern (consisting of either coarse linear or curvilinear opacity or fine subpleural reticulation without substantial ground-glass opacity), mixed pattern (combination of consolidation, ground-glass opacity, and reticular opacity in the presence of architectural distortion), and honeycomb pattern. Furthermore, ground-glass opacity was subcategorized into (a) pure ground-glass opacity, (b) ground-glass opacity with smooth interlobular septal thickening, (c) ground-glass opacity with intralobular lines (crazy-paving pattern), and (d) irregular lines and interfaces with architectural distortion superimposed on ground-glass opacity. Examples of those patterns from our patient cohort are shown in Figure E1 (online). The distribution of abnormalities was also classified as predominantly subpleural (involving mainly the peripheral one-third of the lung), central (involving mainly the central two-third of the lung), or diffuse. In addition, the presence of pleura effusion, pneumothorax, mediastinal emphysema, or mediastinal lymphadenopathy (axial diameter >1.0 cm) was also noted.

We quantified the CT images by using a previously published method (9). Briefly, each lung was divided into upper (above the carina), middle, and lower (below the inferior pulmonary vein) zones. Each zone was evaluated for percentage of lung involvement on a scale of 0 to 4, where 0 = 0% involvement, 1 = less than 25% involvement, 2 = 25% to less than 50% involvement, 3 = 50% to less than 75% involvement, and 4 = 75% or greater involvement. Overall CT scores were the summation of scores from all six lung zones. The maximum possible score was 24.

Statistical Analysis

Illness day 0 was defined as the day of initial symptom onset. The median values of total CT scores and number of zones involved as a function of time were plotted. The temporal changes of main CT patterns, subtypes of ground-glass opacity, and distribution of lung abnormalities were also analyzed. The Kruskal-Wallis rank sum test was used for the difference in the median values of CT lung quantification in different periods, and the χ² test was applied to compare the frequency of CT patterns in different periods. P < .05 was considered to indicate a statistically significant difference. Statistical analyses were performed by using R software (version 3.5.0; R Foundation for Statistical Computing, Vienna, Austria).

Results

Patients

From January 16 to February 17, 2020, 107 patients were admitted to the isolation wards. Among them, 17 patients did not have a positive SARS-CoV-2 RT-PCR result, resulting in a total of 90 patients enrolled in the final study. The median follow-up period was 18 days from admission (range, 5–43 days). The median illness duration from symptom onset to hospital discharge, death, or the end of the study was 25 days (range, 10–56 days).The demographic characteristics, initial symptoms, and clinical outcomes are summarized in the Table.

A total of 366 chest CT scans were acquired in the 90 patients. The numbers of CT scans obtained each day from symptom onset are listed in Figure E2 (online). In 10 patients, their first CT scans were obtained before symptom onset (median time before symptom onset, 5 days; range, 2–9 days), after they had close contact with patients suspected of having or confirmed to have COVID-19. Each patient had a median of four chest CT scans (range, 1–7), with a median scan-to-scan interval of 6 days (range, 2–19 days). Thus, chest CT scans obtained every 6 days after symptom onset were categorized as one group for analysis. Of note, there were only 54 chest CT scans in total for illness days of 24 or more, and these scans were combined and analyzed as one group.

Temporal Changes of CT Scores and Number of Zones Involved

Figure 1 shows the temporal changes of CT lung quantifications. There was a marked increase in the median values of total CT scores and the number of zones involved after symptom onset. Both the median total CT scores and the median number of zones involved peaked during illness days 6–11, with a median total CT score of 5 (interquartile range, 2–6; P < .05, compared with that of illness days 0–5) and a median of five zones involved (interquartile range, 2–6; P < .05). After that, the high levels persisted until illness day 24 or longer, and no significant changes were observed for both the median values of total CT scores and the number of zones involved (P = .31 and .76, respectively). The trends of both quantifications were similar.

Figure 1:

Download as PowerPoint

For total CT scores, the peak values were reached on median illness day 10 (interquartile range, 6–13 days). In 42 of 88 patients (48%, two patients died after the first CT scan), the total CT scores reached the peak values during illness days 6–11 (Fig 1, C).

Temporal Changes of Main CT Patterns

Six of the 10 patients who underwent their first CT examinations before symptom onset had normal CT scans. The other four patients had either ground-glass opacity (n = 2) or consolidation (n = 2). The period from the date of the abnormal CT scan to symptom onset ranged from 2 to 6 days.

The predominant pattern of abnormality after symptom onset was ground-glass opacity, with the percentage varying from 45% (35 of 78 scans) on illness days 12–17 to 62% (49 of 79 scans) on illness days 0–5 (Fig 2, A). Consolidation was the second most prevalent finding during illness days 0–5 and 6–11, with percentages of 23% (18 of 79 scans) and 24% (20 of 85 scans). Of note, 11 of 79 CT scans (14%) obtained during illness days 0–5 and one of the 85 (1%) obtained during illness days 6–11 were normal, leading to an approximate sensitivity of using CT for SARS-CoV-2 infection of 84% (53 of 63 patients; 95% confidence interval [CI]: 73%, 92%) and 99% (74 of 75 patients; 95% CI: 93%, 100%) for each period.

Figure 2:

Download as PowerPoint

The percentage of ground-glass opacity dropped significantly from 62% (49 of 79 scans) on illness days 0–5 to 45% (35 of 78 scans) on illness days 12–17 (P = .046), with a large increase in the percentage of mixed pattern from 1% (one of 79 scans) to 38% (30 of 78 scans). The mixed pattern became the second most predominant pattern thereafter, with the percentage ranging from 22% (12 of 54 scans) on illness days of 24 or more to 38% (30 of 78 scans) on illness days 12–17. The reticular pattern was rarely seen and was only present on illness days 18–23 and day 24 or more, with percentages of 3% (two of 60 scans) and 6% (three of 54 scans), respectively.

Note that the percentage of ground-glass opacity showed a gradual increase from 45% (35 of 78 scans) on illness days 12–17 to 61% (33 of 54 scans) on illness day 24 or higher, although the difference was not statistically significant (P = .10). To verify this upsurge in ground-glass opacity, patients with an illness duration of at least 25 days from symptom onset (the median illness duration) were selected (48 patients with 204 CT scans). The temporal changes of main CT patterns are shown in Figure E3 (online). Note that those patients contributed to 96% of CT scans (52 of 54) on illness day 24 or higher. A gradual increase from 39% (14 of 36 scans) on illness days 12–17 to 60% (31 of 52 scans) on illness day 24 or more was also observed for the percentage of ground-glass opacity, although the difference was not statistically significant (P = .09).

Temporal Changes of Ground-Glass Opacity

Only pure ground-glass opacity could be observed on CT scans before symptom onset (two in two scans). Pure ground-glass opacity was also the most commonly seen subtype of ground-glass opacity after symptom onset, with the percentage ranging from 40% (20 of 50 scans) on illness days 6–11 to 71% (20 of 28 scans) on illness days 18–23 (Fig 2, B). Ground-glass opacity with superimposed intralobular lines (crazy-paving pattern) was the second most common subtype on illness days 0–5, with a percentage of 24% (12 of 49 scans). The percentage of irregular lines and interfaces superimposed on ground-glass opacity increased significantly from 8% (four of 49 scans) on illness days 0–5 to 28% (14 of 50 scans) on illness days 6–11 (P = .02) and became the second most commonly seen pattern of ground-glass opacity thereafter.

Of note, the percentage of pure ground-glass opacity showed a trend of “first falling then rising.” The percentage dropped significantly from 65% (32 of 49 scans) on illness days 0–5 to 40% (20 of 50 scans) on illness days 6–11 (P = .02), with an increase in the percentages of the other three subtypes. As the percentage of pure ground-glass opacity increased significantly to 71% (20 of 28 scans; P = .02, as compared with that of illness days 6–11) on illness days 18–23 and 70% (23 of 33 scans, P = .02) on illness day 24 or higher, those of the other three patterns gradually decreased.

Images in three patients who underwent a series of CT examinations are shown in Figures 3–5.

Figure 3:

Download as PowerPoint

Figure 4:

Download as PowerPoint

Figure 5:

Download as PowerPoint

Temporal Changes of the Distribution of CT Abnormalities

Fifteen of the 90 patients (17%) had lung lesions that were confined to a single lung throughout the course of the disease. Among them, 14 patients (93%) were discharged from the hospital before the end of the study and four discharged patients (27%) had complete resolution of lung abnormalities on the final CT scans. For the remaining 75 patients (303 CT scans), bilateral lesions were present in 257 of the 303 scans (85%). The temporal changes are depicted in Figure E4a (online). We can see that the unilateral involvement appeared only in the early and very last stage of the disease (illness days <0, 0–5, 6–11, and ≥24), and the proportions gradually decreased with time after symptom onset from 27% (18 of 66 scans) on illness days 0–5 to 12% (six of 71 scans) on illness days 6–11. All lesions were bilateral on illness days 12–17 until 18–23.

In the four patients who had lung lesions before symptom onset, three had subpleural lesions (75%) and one had central lesions. The lung abnormalities were predominantly subpleural after symptom onset, with the percentage ranging from 67% (36 of 54 scans) on illness day 24 or higher to 73% (44 of 60 scans) on illness days 18–23 (Fig E4b [online]). The percentage of diffuse pattern increased largely from 13% (10 of 79 scans) on illness days 0–5 to 29% (25 of 85 scans) on illness days 6–11 and remained relatively stable thereafter. The central distribution pattern was rarely seen after symptom onset and was only present on illness days 0–5, 6–11, and 12–17, with percentages of 3% (two of 79 scans), 1% (one of 85 scans), and 1% (one of 78 scans), respectively.

Other CT Findings

Six of 90 patients (7%) had mild bilateral pleural effusion. In three of those six patients, pleural effusion was present during the entire course of the disease. Pleural effusion developed on illness days 11, 21, and 24 in the other three patients and lasted until the last CT scans. No pneumothorax, mediastinal emphysema, or mediastinal lymphadenopathy were observed on CT images.

CT Findings at Hospital Discharge

Seventy patients were discharged from the hospital at the end of the study. The median hospital stay was 16 days (range, 5–37 days), and the median illness days from symptom onset to discharge was 24 days (range, 10–44 days). The final CT scans were obtained at a median of 2 days (range, 0–8 days) before discharge. In four of the 70 patients, the last CT scans showed complete resolution of lung abnormalities. The remaining 66 patients had residual disease on the last CT scans. The median CT score and median number of zones involved for all discharged patients were 4 (interquartile range, 2–6) and four (interquartile range, 2–6), respectively. The distribution of the CT patterns is shown in Figure 6. Ground-glass opacity was the predominant abnormality found on the last CT scans (42 of 70 patients [60%]; 95% CI: 48%, 72%), followed by a mixed pattern (17 of 70 patients [24%]; 95% CI: 15%, 36%). Pure ground-glass opacity was the most common subtype of ground-glass opacity, with a percentage of 74% (31 of 42 patients; 95% CI: 58%, 86%). Ground-glass opacity with irregular lines and interfaces was noted in 19% (eight of 42 patients; 95% CI: 9%, 34%) of all ground-glass opacities.

Figure 6:

Download as PowerPoint

Discussion

In this study, we systematically depicted the longitudinal changes of CT findings in coronavirus disease 2019 pneumonia. Our results suggest that the lung abnormalities increased quickly after the onset of symptoms, peaked around 6–11 days, and were followed by persistence of high levels in extent for a long duration. In addition, the diverse CT manifestations progressed in a specific pattern over time.

The CT findings of COVID-19 pneumonia reflected a typical lung injury of viral pneumonia, which was characterized by a rapid change as seen in severe acute respiratory syndrome and Middle East respiratory syndrome (9–11). In our patient cohort, the peak levels of lung involvement were reached approximately 6–11 days from symptom onset, which was consistent with that of 10 days reported by Pan et al (7) and the time range of within the 2nd week as documented in severe acute respiratory syndrome (9). After that, persistence of high levels for both total CT scores and the zones involved were observed, indicating a slow absorption of COVID-19 pneumonia. The slow absorption was similar to that reported in severe acute respiratory syndrome (9). However, Pan et al (7) showed a relatively faster decrease in CT scores after the peak, which might be because only patients with mild COVID-19 pneumonia were included.

The most common CT finding during the course of COVID-19 pneumonia in our patient cohort was ground-glass opacity alone. Consolidation was the second most common feature seen in the first 11 days. When combined, ground-glass opacity and consolidation constituted about 83%–85% of all CT findings in total in the early stage of the disease. This was consistent with findings in prior reports (5). The CT manifestations became more diverse during illness days 6–11 and later. A mixed pattern with architectural distortion, mainly evolving from ground-glass opacity, became the second most predominant pattern since illness days 12–17. Of note, at least part of the mixed pattern characterized by perilobular abnormalities might suggest the presence of secondary organizing pneumonia, which needed to be confirmed with lung biopsy or bronchoalveolar lavage (12). Secondary organizing pneumonia resulting from viral pneumonia had been documented in severe acute respiratory syndrome coronavirus, Middle East respiratory syndrome coronavirus, and influenza infections (13–15). Given that organizing pneumonia had the potential to progress to fibrosis and corticosteroid therapy was effective in the treatment, follow-up CT scans might be needed for the early identification of patients suspected of having organizing pneumonia in COVID-19, especially those with typical CT manifestations (16,17). An increase in ground-glass opacity had been observed in late stages. The upsurge has also been shown in patients with relatively severe disease with longer illness duration, although the difference was not statistically significant. The lack of significance might be due to the small sample size, and thus needs to be validated in a larger patient cohort. However, the upsurge in ground-glass opacity in late stages had also been seen in severe acute respiratory syndrome (9). It might suggest that ground-glass opacity was the last stage of the illness, and the change of patterns to ground-glass opacity in the late stage showed absorption or recovery of the illness.

For ground-glass opacity, pure ground-glass opacity was the most common subtype during the course of illness. Ground-glass opacity with superimposed irregular lines and interfaces was more commonly seen since illness days 6–11. This pattern could be mainly evolved from pure ground-glass opacity, and the linear opacity might be due to subsegmental atelectasis or secondary organizing pneumonia. The percentage of pure ground-glass opacity gradually and significantly increased from illness days 6–11, which could be accounted for by a decrease in the other three subtypes. This change in pattern probably represents the gradual resolution of inflammation with re-expansion of alveoli and thus indicates the absorption or recovery of illness.

Of note, two approximate sensitivities of 84% (95% CI: 73%, 92%) and 99% (95% CI: 93%, 100%) were estimated for illness days 0–5 and 6–11, respectively. Those results indicated that the sensitivity of CT for SARS-CoV-2 infection increased over time after symptom onset, which was similar to findings in previous reports (18). Several other studies also reported sensitivities, with percentages varying from 44% to 98% (18–21). The difference between those sensitivities might be partly explained by the different duration from symptom onset to inclusion.

Of the 70 patients who were discharged from the hospital at the end of the study, 66 (94%) still had mild to substantial residual lung abnormalities on their last CT scans. The main pattern of those lung abnormalities was ground-glass opacity. A recent article reported that four discharged patients had positive SARS-CoV-2 RT-PCR results again 5–13 days after discharge (22). Thus, follow-up monitoring of patients might still be necessary.

Of note, the data on illness day 24 or more were not always consistent with the trend as observed before illness day 24. A reasonable explanation was that the CT scans on illness days 24 or more were mainly from those patients with a longer illness duration, probably related to more severe illness.

There are several limitations in our study. First, subgroup analysis of patients with mild and severe disease was not performed. Subgroup analysis would also be of great benefit in determining potential prognostic factors. Second, a larger sample size and longer follow-up are needed to better depict the development of the illness.

In conclusion, the most common CT manifestations in coronavirus disease 2019 pneumonia were bilateral ground-glass opacity with subpleural distribution and absence of pleural effusion. The extent of CT abnormalities progressed rapidly after the onset of symptoms, peaked around 6–11 days, and was followed by persistence of high levels. The temporal changes of the diverse CT manifestations followed a specific pattern, which might indicate the progression and recovery of the illness.

Disclosures of Conflicts of Interest: Y.W. disclosed no relevant relationships. C.D. disclosed no relevant relationships. Y.H. disclosed no relevant relationships. C.L. disclosed no relevant relationships. Q.R. disclosed no relevant relationships. X.Z. disclosed no relevant relationships. H.S. disclosed no relevant relationships. M.Z. disclosed no relevant relationships.

Author Contributions

Author contributions: Guarantors of integrity of entire study, H.S., M.Z.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; agrees to ensure any questions related to the work are appropriately resolved, all authors; literature research, Y.H., H.S., M.Z.; clinical studies, Y.W., Q.R., X.Z., H.S., M.Z.; experimental studies, C.L.; statistical analysis, C.D., Y.H., M.Z.; and manuscript editing, Y.W., C.D., Y.H., H.S., M.Z.