ABSTRACT
COVID-19 can present with respiratory symptoms ranging from mild cough to viral pneumonia and ARDS. Lung ultrasonography has emerged as a promising imaging modality during the pandemic, but there is still a paucity of systematic analysis of lung ultrasound findings.
In this retrospective observational study, 12 Zone ultrasound scans of COVID-19 positive patients were systematically analysed for pleural irregularities, subpleural consolidations, B lines, deep consolidations and effusions. Lung abnormalities were analysed according to overall frequency, frequency distribution in coronal and sagittal lung planes and were also correlated to clinical severity groups as determined by oxygenation deficit.
Our results show that lung zones with abnormalities can occur juxtaposed to normal lung. Irregular pleural and small subpleural consolidations appear ubiquitously distributed throughout both lungs and occur early in the disease process. Wide B-lines are a predominant feature in COVID-19 infection. B-lines are found in a variety of patterns with number and width correlated to disease severity. In our analysis we also describe a previously unrecognised finding of small peri-pleural effusions in 8.7% of scans occurring in all areas of the lung.
The current results form the basis for a more thorough understanding of the lung changes occurring in COVID-19 and the incorporation of lung ultrasound in the setting of COVID-19 infection including triage, diagnosis, treatment approach and prognosis.
INTRODUCTION
COVID-19 is the disease caused by SARS-CoV-2 infection, first identified in China in 2019, which spread rapidly across the world with pandemic status declared in March 2020 by the World Health Organisation[1]. Clinically the infection has a broad range of manifestations, but presents predominantly as a respiratory illness with a wide range of severity, including asymptomatic cases, mild flu-like symptoms, as well as viral pneumonia, and in extreme cases ARDS-like illness with a high mortality rate[2].
Evidence from computerised tomography (CT) imaging of the chest showed that COVID-19 infection gives rise to a classic appearance of patchy peripheral ground-glass opacities (GGO) with ill-defined margins and inter- and intralobular septal thickening leading to a so-called “crazy paving” cobble-stone patterns. There are also associated pleural abnormalities[3]. Corresponding changes are demonstrable on lung ultrasound (LUS) rendering it a potentially useful tool in the diagnosis of COVID-19. This is especially so given the poor sensitivity and specificity of CXR for the changes typical for COVID-19[3]. The usefulness of lung ultrasound in the context of viral pneumonias has been previously explored in the H1N1 (swine-flu) and H7N9 (avian influenza) pandemics in 2009 and 2013 respectively, with high sensitivity and specificity[4][5].
LUS allows conclusions about the lung by interpreting artefacts. In normal lung parenchyma, horizontal reverberation artefacts are commonly seen, known as A-lines. Interstitial-alveolar syndrome with increased septal thickening and peri-alveolar oedema results in multiple vertical ring-down artefacts which arise from the pleural line and move with pleural sliding, known as B-lines. B-lines can be seen in multiple disease processes that cause alveolar-interstitial abnormalities including pneumonia, pulmonary oedema, interstitial lung disease and others. Although isolated thin B-lines can be present in lung bases without signifying pathology, more than two B-lines per intercostal space is considered pathological. Multiple B-lines correlate to GGO on CT, and more than six B-lines per intercostal space or the appearance of ‘white lung’ from completely coalesced B-lines indicate severe interstitial syndrome or alveolar oedema[6][7][8].
Peng and colleagues characterised the findings demonstrated in lung ultrasound of COVID-19 positive patients[9]. They described pleural thickening and irregularities, B-lines and consolidations in a variety of patterns, and an absence of large pleural effusions[9]. These findings have been echoed by many in the ultrasound community[10][11][12][13][14][15]. In addition, Volpicelli et al. have identified a characteristic B line appearance they termed “the light beam sign”, seen as a wide band-like artefact sliding with the pleura thereby creating and ‘on-off’ effect[16]. The findings are similar to those described for ARDS showing areas of abnormal lung interspersed by normal, spared lung.
There is currently a paucity of data correlating clinical features of COVID-19 with specific lung sonographic findings and as of yet, the use of lung ultrasound for prognostication of disease severity has not been explored. In this retrospective observational study, we systematically assess abnormalities found on lung ultrasound of COVID-19 positive patients (pleural irregularities, B-lines, consolidations, pleural effusions) using a 12-zone scanning approach. Each abnormality is analysed individually according to frequency in different lung areas in coronal and transverse cross-section. Moreover, the findings are correlated to clinical severity as determined by SpO2/FiO2 (S/F) ratio, with cut-off values based on the Berlin Criteria for the diagnosis of ARDS[17]. To our knowledge, this is the first systematic study of this kind. The findings from our study could guide future research into the utility of lung ultrasound in the diagnosis and prognostication of COVID-19.
METHODS
Selection of patients
This was a retrospective observational study undertaken in a busy urban Emergency Department (ED) in the United Kingdom (UK). Patients who had presented to the ED in a 6-week period (from mid-March to the end of April 2020) with symptoms consistent with COVID-19 as per government guideline [18] and also had received a LUS as part of their clinical assessment, were considered for the study. The LUS was performed by one of two ED doctors trained in LUS. Retrospectively, it was ascertained which of these patients were COVID-19 positive by cross-referencing their case notes on e-Record™ patient. COVID-19 positive status was defined either as positive RT-PCR or combined laboratory results (lymphopaenia) and positive CXR or CT report (reported by Consultant Radiologists blinded to blood results).
Patients were divided into 4 clinical severity groups based on S/F ratio: normal oxygenation, mild (S/F 235-315), moderate (S/F 148-235) and severe (S/F <148) oxygenation deficit. The cut-off values were based on the Berlin Diagnostic criteria for ARDS [17]. The S/F ratios were translated into PaO2/FiO2 (P/F) ratios as suggested by Rice et al [19][20] (Table 2).
Ultrasound
LUS was performed by one of two doctors in the Emergency Department trained in point-of-care ultrasound (POCUS) using a low frequency (3-5 MHz) transducer. Two ultrasound machines were used in the department - a SonoSite Xport ™ (FUJIFILM SonoSite Inc, Bothell, WA, USA) with set to abdominal pre-set with Tissue Harmonic Imaging (THI) and Compound Imaging (CI) manually turned off, or a GE Venue™ (GE Healthcare, Chicago, Illinois, US) with lung pre-set. The intercostal spaces were scanned sagitally, at a depth of 13-16 cm. The focal zone was set to the pleural line. Still images and clips of each zone were saved for analysis.
Twelve lung zones were assessed where the patients’ clinical condition allowed. Zones were identified as follows: each hemithorax was divided by anterior and posterior axillary lines into anterior, lateral and posterior areas, which were further subdivided into superior and inferior zones, resulting in 6 scanning zones per hemithorax. Labelling was set according to hemithorax, R (right) and L (left) and numbered zone 1 and 2 for anterior zones, 3 and 4 for lateral zones, and 5 and 6 for posterior zones, as depicted in Figure 2 (Figure 1).
Review process
Images and clips were anonymised and independently reported by two clinicians with experience in LUS who were blinded to patients’ clinical information. One of the interpreting doctors was an expert with >5 years experience in POCUS and the other had completed a formal ultrasound fellowship. Reporting was carried out using an ultrasound proforma that was specifically developed for assessment of COVID-19 LUS in the department (Figure 1). Images that were ambiguous were excluded. The reports were compared categorically, and inter-rater reliability was assessed using Cohen’s Kappa.
The resulting sonographic clips or images were scored for the presence (positive or negative) of the following ultrasound appearances: A-line pattern (A), pleural irregularity (P1), small peripheral consolidation (<1cm) (C1), larger consolidation (C2), B-lines: ≤ 2 isolated (B1), >2 B-lines or B-lines with a thickness occupying >25% of the pleura at origin (B2), confluent B-lines resulting in a ‘white lung’ appearance (B3), and pleural effusions: small (<1cm) (E1), large (E2) (Figure 2).
Positive ultrasound findings were compared by lung areas. Comparisons were made between upper (Zones 1&5), middle (Zones 2&3), and lower (Zones 4&6) lung areas, as well as between anterior (Zones 1&2), lateral (Zones 3&4), and posterior (Zones 5&6). To assess whether there was a statistically significant difference between the positive ultrasound findings, 95% confidence intervals (with a continuity correction) were calculated for each group. Instances where the 95% confidence intervals do not overlap between compared groups were considered significant.
RESULTS
During the study period, 50 patients who presented with symptoms consistent with COVID-19 disease had a LUS performed. After cross-referencing with case notes, 44 patients were deemed to be COVID-19 positive and included in the study. Patient characteristics are as shown in Table 1.
The patient demographics included 28 males and 16 females, 22 patients over the age of 60 and 26 having prior cardiovascular morbidities. Dyspnoea was the main presenting complaint in 36 patients. Diagnostic PCR test of a throat/nose swab was positive for 35 patients. CXR was reported as diagnostic for COVID-19 for 34 patients and 24 patients were found to be lymphopaenic. Patients were grouped into severity groups according to S/F ratios as described in Table 2, resulting in 18 patients classified as having normal oxygenation, 5 patients as having mild oxygenation deficit, 7 as moderate and 13 as severe. In total, 492 zones were analysed across the 44 patients included in the study.
All scans were reported independently by the two clinicians and were compared for agreement. Reports were categorised as ‘agree’ or ‘disagree’; the inter-rater reliability agreement was 91% with a Cohen’s K value of 0.8.
Prevalence of different lung abnornalities
The LUS appearances found on the 492 scans analysed were A-line pattern (A), pleural irregularities (P1), small (<1 cm) peripheral consolidations (C1), larger consolidations (C2), small (<1cm) peripleural effusions (E1), larger effusions (E2), and a range of B-line patterns from sparsely distributed thin B-lines (B1) to more frequent or wide band-like (B2) and completely coalesced B lines filling out the entire pleural space (B3). The most prevalent findings were P1 and C1, found in 47.5% and 28.8% % of all scans respectively. As many as 47.5% of all scans showed a ‘white lung’ appearance (B3). Abnormal lung was frequently interspersed with areas of normal lung with A-line pattern demonstrated in 27.1% of all scans. B1 and B2 were seen in 5.1% and 11.9% respectively, whereas larger consolidations and larger effusions were seen infrequently, each represented in 6.8% of all scans. We also demonstrated a proportion of small peri-pleural effusions (E1) in 15.3% of all scans. Each lung abnormality was found in isolation, as well as together with other abnormalities described above. The abnormalities were found throughout all lung zones and occurred bilaterally (Figure 3A, Table 3).
Distribution of lung abnormalities amongst all 12 lung zones
The distribution of the types of lung abnormalities between different areas of the lung was analysed comparing 95% confidence intervals. When comparing upper zones (zones 1&5), middle zones (zones 2&3) and lower zones (zones 4&6), there was no statistically significant difference between the proportion of scans showing A-lines. Irregular pleura was present in 53.1% (45.1-60.9), 62.6% (54.8-69.8) and 62.9% (54.8-70.3) of upper, middle, and lower zone scans, respectively. B1 was found in 8.6% (5-14.4) of upper lung zones, 6.4% (3.4-11.5) of middle lung zones and 6.3% (3.2-11.6) of lower lung zones. For the same order of zones, B2 was found in 14.2% (9.4-20.1), 15.8% (10.8-22.3) and 11.9% (7.5-18.3) and B3 was found in 27.8% (21.2-35.5), 26.3% (20-33.7) and 35.8% (28.5-43.9). Small consolidations (C1) were apparent in 43.2 (35.5-51.2), 43.9 (36.4-51.6) and 47.2% (39.3-55.2). No larger consolidations were found in the upper lung zones, whereas they were found in 1.8% (0.5-5.4) of middle zone scans and 1.9% (0.5-58.5) of lower lobes. Small effusions (E1) were found in 9.3% (5.5-15.1) of upper, 2.9% (1.1-7.1) of middle and 14.5% (9.6-21.1) of lower zones with a statistically significant higher occurrence in the lower zones when compared to the middle zones. Large effusions (E2) only found in the lower zones, occurring in 6.9% of lower zone scans which conveys a significant difference (Table 4). There were no significant differences in findings between right and left hemithoraces (Figure 3B&C).
Distribution of lung abnormalities amongst different anatomical planes
When comparing anterior, lateral and posterior areas of the lung, A lines were seen in 35.8% (28.8-43.4) of anterior scans, 24% (17.8-31.3) of lateral scans and 29.5% (22.5-37.6) of posterior scans. Irregular Pleura was present throughout all lung areas with 60.8% (53.1-68), 61.7% (53.8-69) and 55.7% (47.4-63.8) and for anterior, lateral and posterior areas specifically. B1 appearance was apparent in 7.4% (4.2-12.6), 7.2% (3.9-12.5) and 4.7% (2.1-9.8) of scans and B2 seen in 18.2 (12.9-24.9), 13.8 (9.1-20.2) and 9.4% (5.4-15.6) of anterior, lateral and posterior scans respectively. B3 was found in 27.8% (21.5-35.2) of scans in anterior zones, 28.7% (22.1-36.3) of lateral scans and 33.6% (26.2-41.8) of posterior scans. Small consolidations appeared equally distributed throughout all areas with 45.5% (38-53.1), 43.1% (35.6-51), and 45.6% (37.5-54) for anterior, lateral and posterior zones. Large consolidations (C2) appeared in 1.7% (0.4-5.3), 4.2% (1.9-8.8) and 2.0% (0.5-6.2) for the same order of areas. Small effusions were seen in 17.4% (11.9-24.7) of posterior scans, significantly higher than anterior (5.7% (2.9-10.5)) and lateral (4.2% (1.9-8.8)) areas. Large effusions were not demonstrated in anterior zones, but seen in 4.2% (1.9-8.8) of lateral scans and 2.7% (0.8-7.2) of posterior zones (Figure 3D&E, Table 4).
Distribution of lung abnormalities amongst different clinical severity groups
When comparing scan findings between the four clinical severity groups, the percentage of A lines was significantly different between all clinical severity groups, occurring in 55.2% (48.1-61.9) of the ‘satisfactory oxygenation’ group, compared to 27.1% (16.8-40.5), 6.8%(2.8-14.8) and 6.0% (2.8-11.9) in the groups with mild, moderate and severe oxygenation deficiency. Irregular pleura also was found with significant differences amongst the groups, namely 39.2% (32.6-46.1), 47.5% (34.5-60.1), 76.1% (65.7-84.3) and 85.7% (78.3-90.1) for the satisfactory, mild, moderate and severe groups respectively. In the same group order, B1 was found in 4.7%, 5.1%, 13.6% and 5.4%, and B2 was demonstrated in 11.3%, 11.9%, 18.2% and 16.5%, neither demonstrated significant difference. There was a statistically significant difference however, when comparing B3 appearance between scans from patients with satisfactory oxygenation, where it was found in 12.3% (8.3-17.6) compared to the other groups with 47.3% (34.5-60.1), 39.8% (30-50.1) and 39.1% (30.9-48) in mild, moderate and severe groups respectively. Small peripheral consolidations were found significantly less in satisfactory (29.2% (23.3-36)) and mild (28.8% (18.1-42.3)) groups compared to moderate (60.2% (49.2-70.3)) and severe (66.2% (57.4-74)) groups. Large consolidations were not found in the ‘satisfactory oxygenation’ group, but occurred with 6.8% (2.2-17.3) in the mild and 6.8% (2.8-14.8) moderate group, and 2.3% (0.6-7) in the severe group; this represented a significant difference between the patients with satisfactory oxygenation and moderate and severe oxygenation deficit. Small effusions were seen in 6.1%, 15.3%, 12.5% and 7.5% for the groups in increasing severity with no statistical difference, and large effusions (E2) w seen in 0.5%, 6.8%, 4.5% and 1.5% for the same groups, also with no statistically significant difference demonstrated (Figure 4, Table 5). Linear regression analysis showed a strong negative correlation between A lines and clinical severity, a strong positive correlation between pleural irregularity and clinical severity, a positive correlation between B2 and C1 and severity group, and a slight positive correlation between B1 and B3 and severity group (Figure 5).
DISCUSSION
Our comprehensive analysis of 492 lung ultrasound scans of COVID-19 positive patients resulted in a clear pattern of abnormalities seen on LUS in the context of COVID-19 infection. These include pleural irregularities, small peripheral consolidations, a range of B-line patterns from separate thin B-lines to completely coalesced B-lines resulting in the appearance of a white intercostal space, as well as small peri-pleural effusions. Large effusions and consolidations were less prominently featured.
Pleural irregularities and small peripheral consolidations (<1cm) are prevalent in patients affected by COVID-19. They are ubiquitously distributed throughout both lungs and also correlate with clinical severity. There was no clear preference for a transverse (top, middle, bottom) or coronal (anterior/ lateral, posterior) plane. These results appear at first sight somewhat conflicting to reports available in the literature, which suggest a preference of lung abnormalities in the posterobasal lung zones[9][21][12]. However, most of these previous reports did not separate individual types of abnormalities, took into account gravity-dependent large consolidations and effusions, and merely found that altogether there was a higher proportion of abnormalities in the posterobasal areas. In fact, reports from CT scans of affected patients indicate that the patchy areas of ground glass opacification and small consolidations occur bilaterally, peripherally and all lung segments can be involved[22].
In our analysis, we noted the occurrence of small (<1cm) localised effusions which appeared to be situated on the outside of the parietal pleura rather than between the two pleural layers. To date these peri-pleural collections have not been formally described in the literature but have been discussed in expert circles and in this study were seen in 8.7% of all analysed scans. It is conceivable that they represent localised fluid collections as a result of inflammatory reactions. These small effusions can be found in all lung zones defying a gravity dependent distribution, which supports their suspected aetiology as localised reaction.
B-lines can be seen in multiple disease processes that cause alveolar-interstitial abnormalities including pneumonia, pulmonary oedema, interstitial lung disease and others. For the analysis of B-lines in this study, we largely followed leading literature[6][23][8]. To avoid ambiguity, we used a simplified semiquantitative numerical evaluation of B-lines, where B-lines were counted per intercostal space and grouped into B1 (<2 B-lines) and B2 (≥ 2 B-lines). A variety of different B line patterns was found, including separate thin B-lines, multiple coalescing lines, thick confluent bands, as well as the appearance of an entirely white intercostal space. An increase in amount of B-lines in the context of COVID-19 hence could indicate loss of lung aeration and be used to assess disease severity. Thin and scattered B-lines (B1) were incorporated in the analysis to elicit their significance in the context of COVID-19, even though it is known that they are occasionally found in normal patients, especially at the bases. In fact, our results show that thin B-lines are seen very infrequently and do not appear to be a defining feature of COVID-19. A white intercostal space (in our dataset categorised as B3) is a common finding in COVID-19 infection with clear correlation to disease severity as defined by worsening oxygenation (Figure 8). In a different clinical context, including pulmonary oedema and ARDS, this appearance has been postulated to result from a complete confluence of B-lines and has been equated to alveolar-interstitial syndromes, seen as GGO on CT. This indicates that lung disease in COVID-19 shares similarities on LUS with other lung pathologies causing alveolar-interstitial syndromes. The occurrence of wide, broad, band-like B-lines is a prominent observation in this analysis. These wide B-lines seem to play a role in the characterisation of COVID-19 with correlation to disease pathophysiology and severity; an observation that has also been made elsewhere[16]. Hence, in our analysis, wide B-lines were regarded as an abnormal finding, even if they were seen in isolation. For classification purposes, B-lines were scored as wide, if they occupied >25% of the pleural space at origin. In this context, new consensus definitions may be required that take into account wide B-lines considering that their presence may help in differentiating different aetiologies of lung pathology from COVID-19.
Our grouping of disease severity also correlated to the severity of lymphopenia and percentage mortality. In addition, our results also indicate that being male and >60 years of age correlated to disease severity which is consistent with published data[2]. It has also been suggested in the literature that the disease reaches a peak symptomatic severity around day 10, also reflected in our severity groups (Table 1). Pattern, amount and distribution of lung abnormalities as detected on LUS clearly correlates to severity of COVID-19 pneumonia as determined by loss of oxygenation. Hence, it is conceivable to develop a scoring system specifically for LUS findings of COVID-19 patients with an aim to predict disease course or need for ventilation.
With lung changes on LUS correlating to disease severity, a scoring system to predict disease severity or guide therapeutic approaches becomes conceivable. Previously, it has been suggested that the ‘lung ultrasound score’ described elsewhere in the context of cardiogenic pulmonary oedema could be applied[24]. This approach utilises the relationship between the amount of B-lines and loss of lung aeration in ARDS and pulmonary oedema by converting numerically evaluated B lines to a score which can be used to predict disease severity and progression, assess therapeutic effect, or guide ventilation strategies[25][26][27]. This has been validated as a monitoring tool for therapeutic response in cardiogenic pulmonary oedema and to proning in ARDS, however it has not been explored in the context of different disease pathologies such as viral pneumonias. A drawback is the lack of standardisation in the methodology of scoring B-lines and hence a significant reliance on observer interpretation leading to interobserver variability, particularly where B-lines are numerous but not completely confluent. Suggested methods of scoring the amount of B-lines have included semiquantitative methods such numerical evaluation and percentage of pleural space occupied by B-lines, as well as computer algorithms for quantitative scoring[28][29]. This system does not however take into account pleural changes which according to our findings are important, as they correlate to disease severity. Others have suggested to score COVID-19 changes on LUS based on pleural changes[30], but this approach neglects B-lines which as indicated in our results, correlate to loss of lung aeration and are hence an important factor assessing disease severity. We propose, that based on the results described here, any scoring system should negatively weight the presence of A-lines, positively weight pleural irregularities and small subpleural consolidations and broad B-lines.
It is important to remember that the lung changes seen on LUS in COVID-19 patients are not specific to COVID-19 and with the decline in prevalence of this disease caution needs to be taken for potential over- and misdiagnosis. Other causes of viral pneumonia such as H1N1 (swine flu) or H7N9 (avian flu), have shown similar findings on lung ultrasound, including irregular pleura and varying amounts of B-lines[5][31], and further research is needed to investigate whether there are distinguishing factors between different types of viral pneumonias on lung ultrasound. CT findings indicate that viral pneumonias generally are characterised by peripheral, bilateral changes, including subpleural consolidations and GGO with subtle differences seen between pneumonia caused by different viral families[32].
We are aware, that there are limitations with our classification of disease severity, primarily that the Berlin criteria classify severity based on PO2, measured at a PEEP of 5[17]. In this study we analysed a patient population representative of the entire range of lung disease seen in COVID-19, from mild infection with relatively normal oxygen saturation as measured by pulse oximetry to severe ARDS as seen by S/F and P/F ratios. Patients were retrospectively grouped into different disease severity groups using the Berlin ARDS criteria as a guide. As not all our patients with mild symptoms were subjected to arterial blood gas analysis, we refer to research conducted by Rice and colleagues which showed that S/F can be reliably related and converted to P/F ratio[19]. This has also been acknowledged by the authors of the original Berlin criteria[20]. We deliberately do not equate our groups to ARDS severity, but rather use this as a guide to classify severity.
We also acknowledge that the sample size was small but given the unique situation with a fast-evolving pandemic we analysed important lung ultrasound characteristics in the limited time available.
In summary, in this observational study we have characterised the changes seen on LUS in patients diagnosed with COVID-19, and systematically analysed their prevalence individually with relation to lung area, as well as correlated them to clinical severity. Pleural irregularities and small peripheral consolidations occurred in all severities of COVID-19 infection ubiquitously distributed throughout both lungs. Wide B-lines were a prominent feature in COVID-19 infection with correlation to disease severity. We also characterised a novel observation of small peripleural effusions. Areas with abnormal lung findings were interspersed with normal A-line pattern. We demonstrated that LUS abnormalities correlate to disease severity as defined by oxygenation deficit. A strong negative correlation was found between A lines and disease severity and strongly positive correlation between pleural irregularities, small peripheral consolidations and large consolidations. In addition, we demonstrated that the appearance of white lung strongly correlates to the loss of oxygenation.
Our results suggest that irregular pleura and small consolidations represent changes that occur early on in the disease process as they are found in all groups of disease severity, whereas increasingly thick and coalescing B-lines indicate a worsening disease process by developing interstitial-alveolar syndrome with increasing loss of oxygenation. This information could be utilised in many ways, starting from triage scenarios, reaching a timely and accurate differential diagnosis, as well as prognosticating further developments in the disease process.
As COVID-19 pandemic progresses, it would be imperative to investigate how long the lung changes last and whether lung ultrasound could be useful in disease follow-up.
Data Availability
all data is provided in the article