Chest CT Findings in Hospitalized Patients with SARS-CoV-2: Delta versus Omicron Variants.

Background CT manifestations of severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) may differ among variants. Purpose To compare the chest CT findings of SARS-CoV-2 between the Delta and Omicron variants. Materials and Methods This retrospective study collected consecutive baseline chest CT images of hospitalized patients with SARS-CoV-2 from a secondary referral hospital when the Delta and Omicron variants predominated. Two radiologists categorized CT images based on the Radiological Society of North America classification system for coronavirus disease 2019 (COVID-19) and visually graded pneumonia extent. Pneumonia, pleural effusion, and intrapulmonary vessels were segmented and quantified on CT images using a priori developed neural networks, followed by reader confirmation. Multivariable logistic and linear regression analyses were performed to examine the associations between the variants and CT category, distribution, severity, and peripheral vascularity. Results In total, 88 patients with the Delta (mean age, 67 years±15; 46 men) and 88 patients with the Omicron (mean age, 62 years±19; 51 men) variants were included. Omicron was associated with a less frequent typical peripheral, bilateral ground-glass opacity (32% [28/88] versus 57% [50/88]; P=.001), more frequent peri-bronchovascular predilection (38% [25/66] versus 7% [5/71]; P<.001), lower visual pneumonia extent (5.4±6.0 versus 7.7±6.6; P=.02), similar pneumonia volume (5%±10 versus 7%±11; P=.14), and a higher proportion of vessels with a cross-sectional area smaller than 5 mm2 relative to the total pulmonary blood volume (BV5%; 48%±11 versus 44%±8; P=.004). In adjusted analyses, Omicron was associated with a non-typical appearance (odds ratio, 0.34; P=.006), peri-bronchovascular predilection (odds ratio, 9.2; P<.001), and higher BV5% (β value, 3.8; P=.01) but similar visual pneumonia extent (P=.17) and pneumonia volume (P=.67) relative to Delta variant. Conclusions On chest CT, the Omicron SARS-COV-2 variant showed nontypical, peri-bronchovascular pneumonia and less pulmonary vascular involvement than the Delta variant in hospitalized patients with comparable CT disease severity.

Summary statement

The Omicron SARS-CoV2 variant showed more frequent nontypical CT findings (peri-bronchovascular predilection, less pulmonary vascular involvement) than the Delta variant in hospitalized patients with COVID-19 disease with comparable CT severity.

■ Only 32% of patients with the Omicron SARS-CoV-2 variant had typical ground-class opacity at CT versus 57% of those with Delta variant (P=.001);

■ Peri-bronchovascular predilection was greater for Omicron versus Delta variant (38% vs 7%; P <.001, respectively).

■ Pneumonia extent (P =.17) and volume (P =.67) were not different between the variants after adjusting for confounders of age, comorbidities, vaccination, and infection duration.

Introduction

Since the coronavirus disease 2019 (COVID-19) pandemic began, severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) has evolved through genetic mutations during viral replication (1, 2). Several variants of SARS-CoV-2 have been reported worldwide throughout the pandemic, and some variants are classified as variants of concern due to increased transmissibility, disease severity, or evasiveness of treatments and vaccines (3). The Delta and Omicron variants are the two latest variants of concern (4, 5).

The Delta variant was first identified in India in December 2020 and became the globally dominant strain in June 2021. This variant of concern has mutations that make it highly transmissible (more than 60% higher transmissibility than the previous variant), less responsive to antibodies and treatment (i.e., reduced neutralization by antibodies generated against previous infection or vaccination), and more likely to cause adverse outcomes (e.g., severe cases, hospitalization, deaths) (6). Indeed, the Delta variant caused the second wave of India's pandemic and subsequent waves in other countries (7). The Omicron variant, which was first reported in November 2021 in South Africa, was designated as another variant of concern and has become the dominant strain following the Delta variant in most countries. Although the Omicron variant has up to 3.7 times higher transmissibility than the Delta variant, it is regarded as less virulent in terms of the rate of hospitalization, intensive care unit admissions, and mortality (8-10).

Chest CT plays a key role in the diagnosis, detection of complications, and potential prognostication of patients with COVID-19 (11, 12). Prior studies have investigated differences between these two contiguously emerging dominant variants with a focus on their spike proteins, diagnostic tests, clinical characteristics, transmissibility, and outcomes (9, 10, 13). However, it remains underexplored whether the CT findings of COVID-19 differ among variants. This study aimed to compare the chest CT findings of COVID-19 between the Delta and Omicron variants.

Materials and Methods

The institutional review board approved this retrospective study and waived informed consent (IRB No. SGPAIK 2022-03-009).

Study sample

This study was conducted at one of the secondary referral hospitals for the treatment of patients with mild-to-moderate COVID-19, which operated thirty beds for COVID-19. Inclusion criteria corresponded to patients with a) polymerase chain reaction assay-proven SARS-CoV-2 AND b) mild (no requirement for oxygen treatment) to moderate (a necessity for oxygen treatment with nasal prong or facial mask) severity at admission AND c) any following risk factors for progression (Appendix E1). Exclusion criteria corresponded to a) patients necessitating intensive care unit care or ventilator support at admission, b) pregnant women, and c) patients who chose not to undergo CT scanning.

The hospital routinely performed baseline CT images for hospitalized patients. We collected consecutive baseline CT images in November 2021 and February 2022 (Fig E1) when the Delta and Omicron variants predominated in Korea, accounting for 99% and 97% of cases, respectively (14). All chest CT scans were obtained at full inspiration using one of the following 24- or higher-channel CT scanners (Appendix E1).

Figure E1:

A nationwide trend of predominant variants of SARS-CoV-2 between September 2021 and February 2022.

Since the pandemic, Korea has been conducting genomic surveillance in five representative laboratory centers that geographically cover the entire nation to monitor predominant variants of COVID-19. The centers run whole-genome sequencing and sequencing S protein receptor-binding domain and update the trend of predominant variants weekly. Test specimens approximately cover one-fifths of all RT-PCR positive specimens for SARS-CoV-2. The Delta variant started to be replaced by the Omicron variant from December 2021 through January 2022.

We collected clinico-laboratory information, including on infection duration at the time of the CT scan (i.e., days from symptom onset to the CT scan) (15). The composite outcome was the occurrence of any following events: oxygen ventilation, intensive care unit admission, and mortality.

Visual CT analysis

The randomly assigned baseline CT examinations were independently evaluated by two board-certified thoracic radiologists (S.H.Y. and J.H.L., with 17 and 10 years of clinical experience in thoracic imaging, respectively). They were blinded to the patient's clinical information, including the variant and dates of CT examinations, other than the fact that the patient was infected with SARS-CoV-2. Disagreement between two radiologists were resolved by consensus.

COVID-19 pneumonia CT images were classified into the four categories according to the Radiological Society of North America (RSNA) Expert Consensus Document (16): typical appearance, indeterminate appearance, atypical appearance, negative for COVID-19 pneumonia, developed during the first wave of SARS-CoV2 (Appendix E1). For example, typical features of COVID-19 pneumonia typically include ground glass opacities (GGO) with or without consolidation in a peripheral, posterior, and diffuse or lower lung zone distribution, GGO with round morphology or a “crazy paving” pattern. Bronchial wall thickening, mucoid impactions, and nodules (“tree-in-bud” and centrilobular) seen commonly in infections, are not typically observed.

Pneumonia extent was visually assessed using a scale of 0 to 5 for each of the five lung lobes (17): 0, indicating no involvement; 1, less than 5% involvement; 2, 5%–25% involvement; 3, 26%–49% involvement; 4, 50%–75% involvement; and 5, more than 75% involvement. The total CT score was the sum of the individual lobar scores and ranged from 0 (no involvement) to 25 (maximum involvement). The radiologists also assessed pneumonia density, predilected distribution (bronchovascular versus subpleural), lymphadenopathy, and pleural effusion (Appendix E1).

Quantitative CT analysis

The CT images were processed using commercially available segmentation software (MEDIP PRO v2.0.0.0; MEDICALIP, Seoul, Korea) using a priori developed deep neural networks for segmenting the lung (18), COVID-19 pneumonia (19), pulmonary lobes and fissures, pulmonary vessels (20), and pleural effusion (https://cris.nih.go.kr/cris/search/detailSearch.do/20687): The networks were updated with 3DnnU-Net (21), and the dice similarity scores for those structures were 0.99 (lung), 0.84 (COVID-19), 0.98 (lobes), 0.91 (vessels), and 0.90 (effusion) in internal datasets.

A chest radiologist (S.H.Y.) reviewed and confirmed the segmentation masks. If any corrections were required, an imaging technician manually adjusted the masks under the instruction of the radiologist. The radiologist and technician were blinded to any clinical information other than the fact that the patients were infected with COVID-19. The volume (mL) of the segmented lung parenchymal and pneumonia masks was quantified to determine the proportion of COVID-19 pneumonia in the entire lung parenchyma and each lobe. The mean CT attenuation of COVID-19 was also calculated and converted into pneumonia weight (grams) using an equation based on the CT attenuation and pneumonia volume (22). BV5% was calculated as the percentage of blood volume in intrapulmonary vessels with a cross-sectional area <5 mm2 relative to the total pulmonary blood volume (20). Lower BV5% reflected endothelial dysfunction and loss of microvasculature in COVID-19 (23, 24). The volume of pleural effusion was quantified in milliliters if present.

Statistical analysis

Categorical variables were compared using the Fisher's exact and the χ2 tests, and continuous variables were compared using the t test or Mann-Whitney U test. Inter-reader agreements for visual assessment and total CT score were evaluated using Cohen's kappa coefficient (κ) and intraclass correlation coefficient, respectively.

We examined the correlation between visual CT extent and pneumonia volume using the Pearson correlation coefficient. Multivariable logistic regression analysis was performed to examine the association of the variants with CT category and peri-bronchovascular predilection. Multivariable linear regression analyses were performed to examine the relationships between the variants and visual CT extent, pneumonia percentage, weight with the same adjustment. Multivariable Cox regression analyses were conducted to evaluate the association between the variants and the composite outcome. All multivariable analyses were conducted using an input function of confounders (Appendix E1). Infection duration was classified into five categories: 1, pre-symptomatic; 2, 0-2 days; 3, 3-5 days; 4, 6-11 days; 5, >11 days (15).

The statistical analyses were conducted using SPSS software (version 25.0, IBM), and a two-sided P-value <.05 was considered statistical significance.

Results

Clinical Characteristics of the Study Sample

Of 187 hospitalized patients with SARS-CoV-2, after 11 patients were excluded due to pregnancy (n=8) and reluctance for CT examinations (n=3), the final study sampled was 176. Of the 176 patients, 88 and other 88 patients had the Delta (46 men and 42 women; mean age, 67 years ± 15 [standard deviation]) and Omicron variants (51 men and 37 women; mean age, 62 years ± 19), respectively. All patients denied a prior infection with SARS-CoV-2. A flow diagram is given in Fig 1, and the clinical characteristics of this study sample are described in Table 1.

Figure 1:

Flowchart of patient inclusion. RT-PCR: real-time reverse-transcription polymerase chain reaction assay.

Table 1:

Clinical Characteristics of Patients with COVID-19 according to the Variants

Patients with the Omicron variant had higher levels of vaccination than those with the Delta variant (unvaccinated: 28% [25 of 88] versus 32% [28 of 88]; partially vaccinated: 5% [4 of 88] versus 6% [5 of 88]; fully vaccinated: 23% [20 of 88] versus 62% [55 of 88]; booster vaccinated: 44% [39 of 88] versus 0% [0 of 88]; P<.001). The interval between symptom onset and CT scans was shorter in patients with the Omicron variant than those with the Delta variant (3.9 days ± 3.2 versus 5.5 days ± 4.5; P=.01) in symptomatic patients. Other clinical characteristics, including age, sex, comorbidities, the proportion of asymptomatic patients at the time of CT scans, white blood cell count, lymphocyte count, lactate dehydrogenase, the proportion of patients requiring oxygen treatment, and the proportion of unfavorable outcomes did not show evidence of a difference between the two variants (all P-values >.05) (Table 1).

Visual Assessment and Quantitative Assessment

The CT findings of the Omicron and Delta variants are described in Table 2. In terms of the RSNA COVID-19 imaging classification, the Omicron and Delta variants had different appearances (the proportions of typical, indeterminate, atypical appearance, and negative for pneumonia were 32% [28 of 88], 31% [27 of 88], 13% [11 of 88], and 25% [22 of 88] in patients with the Omicron variant; and 57% [50 of 88], 20% [18 of 88], 3% [3 of 88], and 19% [17 of 88] in patients with the Delta variant; P=.004). When dichotomizing appearance as typical or non-typical, the Omicron variant presented the typical CT appearance of COVID-19 pneumonia less frequently than the Delta variant (32% [28 of 88] versus 57% [50 of 88]; P=.001). The visual score of pneumonia extent was lower in patients with the Omicron variant (mean score, 5.4±6.0) than in those with the Delta variant (mean score, 7.7±6.6; P=.02). However, no evidence of differences were found between the two variants in the CT findings of visual assessment of pneumonia density (predominant GGO, predominant consolidation, and mixed pattern: 68% [45 of 66], 12% [8 of 66], and 20% [13 of 66], respectively, in the Omicron variant versus 66% [47 of 71], 8% [6 of 71], and 25% [18 of 71], respectively, in the Delta variant; P=.61), the presence of lymphadenopathy (11% [10 of 88] versus 15% [13 of 88]; P=.66), and pleural effusion (18% [16 of 88] versus 22% [19 of 88]; P=.71).

Table 2:

Findings on Chest CT Scans of Patients with COVID-19, According to Variant

Regarding inter-reader agreement for visual assessment of CT findings of the two variants, Cohen's kappa coefficients ranged from 0.51 to 0.81 with the highest value of 0.81 (95% confidence interval [CI]: 0.74, 0.88) for RSNA COVID-19 imaging classification and the lowest value of 0.51 (95% CI: 0.41, 0.61) for pneumonia density. The intraclass correlation coefficient for pneumonia extent was 0.98 (95% CI: 0.98, 0.99).

In the quantitative CT analysis, patients with the Omicron variant had a higher BV5% than those with the Delta variant (48% ± 11 versus 44% ± 8; P=.004). The mean CT value (−404 HU ± 139 versus −402 HU ± 115; P=.92), quantitative analysis of pneumonia extent (5% ± 10 versus 7% ± 11; P=.14), pneumonia weight (95g ± 174 versus 143g ± 191; P=.09), and pleural effusion amount (275 mL ± 407 versus 149 mL ± 264; P=.47) showed no evidence of differences between the two variants (Fig 2, Fig 3, Fig 4, and Fig 5). The Pearson correlation coefficient between visual pneumonia extent and pneumonia volume was .84.

Figure 2:

Chest CT from a 66-year-old woman with the Delta variant of COVID-19 with a typical CT appearance.

Unenhanced axial CT images (A, B) show peripheral bilateral ground-glass opacities with some intralobular lines predominantly involving both lower lobes. Segmentation overlay images (C, D) show the segmentation results of pneumonia (red), lobes (orange to violet), and pulmonary vessels with a cross-sectional area <5 mm2 (yellow) and ≥5 mm2 (blue). The visual CT score was 12 points, and the pneumonia volume was 9%.

Figure 3:

Chest CT from a 77-year-old male with the Omicron variant of COVID-19 with indeterminate CT appearance.

Unenhanced axial CT images (A, B) show unilateral peri-bronchovascular ground-glass opacities without intralobular lines or an apicobasal predilection. Segmentation overlay images (C, D) show the segmentation results of pneumonia (red), lobes (orange to violet), effusion (light green), and pulmonary vessels with a cross-sectional area <5 mm2 (yellow) and ≥5 mm2 (blue). The visual CT score was 13 points, and the pneumonia volume was 8%. A focal ground-glass opacity in the lateral portion of left bottom image (D) was not included in the pneumonia mask as it was the minor fissure between right upper and middle lobes.

Figure 4:

Chest CT from an 88-year-old woman with the Delta variant.

A representative three-dimensional image shows lower-lobe predominant pneumonia (pneumonia volume, 14.7%) and a lower percentage of blood volume in intrapulmonary vessels with a cross-sectional area <5 mm2 relative to the total pulmonary blood volume (34.6%). The blue vessels have a cross-sectional area ≥5 mm2, while the yellow vessels have a cross-sectional area <5 mm2. The red indicates COVID-19 pneumonia.

Figure 5:

Chest CT from a 52-year-old male with the Omicron variant.

A representative three-dimensional image shows pneumonia evenly affecting lungs (pneumonia volume, 17.5%) and a preserved percentage of blood volume in intrapulmonary vessels with a cross-sectional area <5 mm2 relative to the total pulmonary blood volume (51.5%). The blue vessels have a cross-sectional area ≥5 mm2, while the yellow vessels have a cross-sectional area <5 mm2. The red indicates COVID-19 pneumonia.

Univariable and Multivariable Analyses for Impact of Omicron Compared to Delta variant

In the univariable analyses, the Omicron variant had a lower frequency of a typical CT appearance (odds ratio, 0.36; P=.001), a more frequent peri-bronchovascular predilection (odds ratio, 8.0; P<.001), a lower visual pneumonia extent (β value, −2.3; P=.02), and greater BV5% (β value, 4.4; P=.004) than the Delta variant, whereas pneumonia volume (β value, −2.3; P=.14), pneumonia weight (β value, −47; P=.17). 30-day composite outcomes were comparable between the variants (hazard ratio, 1.8; P=.21) (Table 3).

Table 3:

Impact of Omicron Variant Compared to Delta Variant on RSNA CT Classification System, CT Severity, Peripheral Vascularity, and the Composite Outcome

The multivariable analyses after adjusting for confounders confirmed that Omicron had a less frequent typical CT appearance (odds ratio, 0.34; 95% CI: 0.16, 0.74; P=.006), a more frequent peri-bronchovascular predilection (odds ratio, 9.2; 95% CI: 2.9, 28; P<.001), and a greater BV5% (β value, 3.8; 95% CI: 0.92, 6.8; P=.01) relative to Delta variant. After adjustment, there were no evidence of differences between the Omicron and Delta variants regarding visual pneumonia extent (β value, −1.09; P=.17), pneumonia volume (β value, −0.62; P=.67), pneumonia weight (β value, −6.9; P=.82), and 30-day composite outcomes (hazard ratio, 3.1; P=.11; hazard ratio: 2.4; P=.29).

Statistically significant confounders were identified in the multivariable analyses: the proportion of patients with a typical CT appearance also was greater with age (P=.006) and an infection duration of six days or longer (P<.001 to .004), while it was lower with full vaccination status (P=.006). The visual extent, pneumonia volume, and pneumonia weight also was greater with age (P<.001 to .009) and a longer infection duration (all P-values <.001), but was lower with full vaccination (P<.001 to 0.08). The BV5% was lower as pneumonia volume was greater (P=.001). CT severity was predictive of developing the 30-day composite outcome regardless of whether it was assessed visually or quantitatively (P=.002 for both).

Discussion

The CT manifestations of COVID-19 among different variants remains underexplored. We found that Omicron variant was associated with a smaller proportion of patients with a typical CT appearance (32% [28 of 88] versus 57% [50 of 88]; P=.001), a larger proportion of patients with peri-bronchovascular pneumonia (38% [25/66] versus 7% [5/71]; P<.001), a lower visual pneumonia extent (5.4 ± 6.0 versus 7.7 ± 6.6; P=.02), similar pneumonia volume (5% ± 10 versus 7% ± 11; P=.14), and a higher proportion of vessels with a cross-sectional area smaller than 5 mm2 relative to the total pulmonary blood volume (BV5%; 48% ± 11 versus 44% ± 8; P=.004). When adjusted for confounders including age, comorbidities, vaccination, and infection duration, the Omicron variant was associated with a non-typical appearance (odds ratio, 0.34; P=.006), peri-bronchovascular predilection (odds ratio, 9.2; P<.001) and higher BV5 (β value, 3.8; P=.01) but not with visual pneumonia extent (β value, −1.09; P=.17) or pneumonia volume (β value, −0.62; P=.67).

The frequency of a typical CT appearance has been reported to vary from 17% to 53% depending on the site and clinical indication (25-28), but our study sample with both variants underwent chest CT at the same site and for the same indications. Similar proportions of patients with both variants were unvaccinated or partially vaccinated, and these patients might be more likely to have a typical CT appearance (29). The Omicron's odds of manifesting with a typical CT appearance and peri-bronchovascular predilection remained significant, even after adjusting for those confounders. The Omicron variant replicates better in the bronchi but worse in the lung parenchyma (13), and these characteristics may hinder Omicron from presenting a typical CT appearance when pneumonia is established in the lung parenchyma, while promoting peri-bronchovascular predilection.

BV5%, which reflects peripheral pulmonary volume and accounts for the majority of pulmonary blood volume (30), has been found to be lower in patients with SARS-CoV-2 than in healthy individuals and patients with acute respiratory distress syndrome (23, 24). Furthermore, a lower BV5% was identified as a predictor of adverse clinical outcomes in COVID-19 (31). This characteristic reduction of BV5% in COVID-19 could result from SARS-CoV-2-induced vasoconstriction or microthrombi of small-caliber vessels (24). Indeed, SARS-CoV-2 pathologically inflames small vessels, provokes thrombi (32, 33), and leads to frequent in-situ pulmonary thrombosis, especially in severe cases (34). Interestingly, early observations suggested that the Omicron variant might have a lower thrombosis rate than previous variants (35). This potentially provides support for the possibility that Omicron might involve fewer pulmonary vessels, in line with there being less involvement of the lower respiratory system.

Lower BV5% and CT vascular engorgement seem to result from the same vascular pathologies of COVID-19 but manifest at different pulmonary vascular calibers. BV5%'s cross-sectional vessel area corresponds to a vascular diameter smaller than 1.26 mm given the equation for a circle. These peripheral minute vessels had vasoconstriction or thrombosed but are too small to be visually assessed on CT. Meanwhile, vascular engorgement in COVID-19 was typically observed in segmental or subsegmental vessels. The diameters were 3-4 mm or larger, corresponding to a cross-sectional vessel area 28-50 mm2 or larger. Vascular engorgement can reflect vascular dilatation or thrombosis proximal to SARS-CoV2-affected microvessels. Taken together, modern CT provided a multi-level analysis for revealing pulmonary vascular pathology in COVID-19 from impaired perfusion (below millimeter), lower peripheral BV5% (around millimeter), and proximally engorged vascular changes (over millimeter).

Our study had limitations. This study was retrospective and only included a relatively small number of hospitalized patients. In addition, patient inclusion was conducted without calculating the sample size. Participants did not undergo testing to confirm the SARS-CoV-2 variant. Third, the clinical severity of hospitalized patients might not have been identical between variants, and the number of COVID-19 cases remained low in November 2021 (when the Delta variant predominated), but soared in February 2022 (Omicron variant). Fourth, the pulmonary vessels could not be segmented in dense consolidation areas on non-contrast CT images as the neural network and radiologist could not trace the vessels within consolidations. Fifth, we did not adjust for the multiplicity of tests in our analyses.

In conclusion, the Omicron variant showed more frequent non-typical, peri-bronchovascular pneumonia and less pulmonary vascular involvement than the Delta variant in hospitalized patients with comparable CT severity. The CT characteristics of Omicron may hamper radiologists from promptly recognizing COVID-19 on CT images when incidentally encountered, and this finding raises an alarm regarding the need to evaluate whether CT findings remain consistent or change when new variants appear.