Imaging of lung pathology in COVID-19 (literature review and own data) изменений в легких при коронавирусной (обзор литературы и собственные данные)

Novel coronavirus infection is predominantly manifests as lung tissue damage. Imaging methods, particularly, chest X-Ray and computed tomography, are of great importance for detecting pulmonary changes and differentiate them with other diseases (mainly other viral pneumonias). In the early disease stages the disease presents on CT with ground glass opacities, consolidations, crazy paving symptom. With time course, they can gradually decrease, evolve into organizing pneumonia or stay stable and even increase in volume with the spread of consolidation and formation of several signs of organizing pneumonia. Although radiological methods show high sensitivity in the detection of pulmonary changes, their specificity and prognostic ability are not so good today. Novel coronavirus infection can be complicated with pulmonary embolism, development thrombosis in situ in pulmonary small vessels, acute heart failure and subsequent development of cardiogenic pulmonary edema, bacterial superinfection, exacerbation or worsening of chronic lung disease and several iatrogenic issues (pneumothorax, pneumomediastinum, hematomas).

The world has been affected by the novel coronavirus infection  since December 2019, when the coronavirus SARS-CoV-2 was first isolated from patients in the Chinese city of Wuhan. Since then, the number of new cases worldwide has been increasing steadily. Lung damage is the main manifestation of this infection, which is seen almost in all cases and defined as "viral pneumonia". The need is clear for timely and accurate assessment of abnormal chest findings, appropriate assessment of these changes over time, and evaluation of the prognostic value of the main radiographic signs of organ damage. Imaging techniques are particularly important for detecting pulmonary changes caused by SARS-CoV-2, differentiating them from those caused by other disorders, assessing the severity of abnormal findings and changes in them over time, and evaluating the efficacy of treatment.

Imaging techniques
Imaging techniques used to detect pulmonary abnormalities in patients with COVID-19 include plain chest X-ray and chest computed tomography (CT), and, in some cases, pulmonary and pleural ultrasound. Magnetic resonance imaging and radionuclide imaging have no diagnostic significance for acute respiratory infections and are not used to detect lung damage in COVID- 19.
During the first days of the disease, the sensitivity of routine X-ray for early changes in the lungs (ground-glass opacities) is relatively low, and this technique cannot be used to exclude COVID-19 during this period [1,2]. At later stages, however, the diagnostic value of X-ray significantly increases as the disease becomes more generalised and pulmonary consolidation progresses. This method reliably detects severe viral lung disease and pulmonary oedema of various etiologies. Portable chest X-ray obtained using mobile units is the main diagnostic imaging modality for chest abnormalities in resuscitation and intensive care units. This type of X-ray is recommended in most clinical guidelines as the most readily available and epidemiologically safe imaging technique.
Computed tomography is the most sensitive tool to detect radiographic signs of COVID-19. CT is feasible for initial chest evaluation in patients with severe, progressive disease and patients with clinical signs of acute viral respiratory infections and evident risk factors for a severe course of infection. Also CT is necessary for differential diagnosis of the findings revealed and, in some cases, for follow-up evaluation. The main drawback of CT is its low specificity, which rarely exceeds 60%, as it reveals multiple unequivocal lesions, such as small ground-glass opacities or reticular lesions. The specificity of CT also depends on the prevalence of infection in a particular area [1,[3][4][5][6]. The higher percentage of people are infected, the more specific CT becomes, and vice versa. The specificity of CT scanning is also highly dependent on the probability of the disease in an individual patient. The lower the clinical probability is, the less specific CT findings are.
Due to their low specificity, none of imaging modalities is recommended as a screening tool to detect respiratory problems in asymptomatic persons and patients with mild disease. In addition, imaging techniques cannot identify the etiology of the disease and, therefore, cannot replace routine laboratory tests.
Ultrasonography is an adjunctive imaging tool, which does not replace or exclude CT scanning and X-ray. When its technique is well respected and the examination is performed by a qualified specialist, ultrasonography is highly sensitive for the detection of subpleural interstitial changes and consolidations in lung tissue, pleural effusion, and pneumothorax [7][8][9].
The role of ultrasound in assessing pulmonary abnormalities in COVID-19 patients is controversial. Despite their high sensitivity, ultrasound data are not helpful in assessing the actual extent of lung injury. Ultrasound findings do not always correlate with X-ray and CT data. Finally, ultrasonography is not a standard diagnostic tool to diagnose pneumonia and not included in clinical guidelines for the diagnosis and treatment of communi-ty-acquired pneumonia. Therefore, its diagnostic value depends largely on the experience and qualifications of physicians in a particular facility.

Comparison of radiographic morphological features
It has been established that SARS-CoV-2 binds to angiotensin-converting enzyme 2 (ACE2), a membrane receptor expressed mainly on the surface of nasal epithelial cells, lower airway cells, particularly type II alveolar cells, as well as cells in the upper third of the oesophagus, enterocytes of the colon, cholangiocytes, cardiomyocytes, and epithelial cells in the proximal renal tubules and the bladder. Therefore, the airway epithelium serves as the standard entry point for this infection; from there the virus moves to the blood and affects the most susceptible organs and tissues [11][12][13].
Viral exposure in the lungs usually triggers diffuse alveolar damage (DAD). This term is used to describe characteristic changes in all layers of the alveolar-capillary membrane, including its basal membrane [13][14][15][16][17]. DAD typically is a biphasic process, and includes exudative (oedematous) and proliferative phases.
The exudative phase develops within the first few days and manifests as inflammation in the area damaged by the virus. This is accompanied by breaches in alveolar epithelium and, in some cases, sloughing of alveolar epithelial cells from the basal membrane. Some histochemical reactions and the loss of the integrity of the alveolar-capillary barrier result in exacerbation of interstitial oedema and filling of the alveoli with a fluid rich in proteins, particularly fibrin. An important step of this process is the formation of hyaline membranes, one of the main morphological markers of DAD, in the alveolar spaces.
Partial filling of the alveoli with exudate, cellular debris, and hyaline membranes is responsible for groundglass opacities, the earliest and the most typical sign of viral lung damage detected by various imaging modalities. As abnormal contents continues to accumulate in the alveolar spaces, images start to show consolidations, areas of completely airless lung parenchyma, which are usually surrounded by a peripheral ring of ground-glass opacity (halo sign).
Viral damage to the alveolar-capillary membrane may spread throughout the interstitial compartment and affect pulmonary capillary endothelial cells. Breakdown of capillary endothelial lining cells and those anchored to the basal membrane results in the following two consequences: damage to the capillary wall (basal membrane) and an outpouring of haemorrhagic exudate directly into the alveolar spaces and the interstitial compartment, on the one hand, and thrombosis of small pulmonary vessels [17].
Radiographically, accumulation of haemorrhagic exudate in alveoli also manifests as ground-glass opacities, which relatively rapidly develop into consolidation. Ground-glass opacity lesions often have areas of reticular abnormality seen as polygonal structures 5 to 15 mm in diameter, which are actually thickened intralobular or interlobular septa (Figure 1). This is the well-known crazy paving sign. It is not specific for viral lung damage. It can also be seen in a number of other infections, as well as in neoplasms, interstitial disorders, pulmonary oedema of various etiologies, pulmonary haemorrhage, and in case of blood aspiration in patients with pulmonary bleeding [18][19][20]. Some studies showed that the crazy paving sign predicts a poor prognosis for coronavirus infection, which has been reflected in many recent guidelines [5].
On radiographic images, bleeding into the alveoli or blood accumulation in the alveolar spaces, on the one hand, and cellular debris with hyaline membranes in the alveoli, on the other, manifest identically, i.e. as groundglass opacity and consolidation. Radiography does not provide reliable evidence to distinguish between these morphological changes. However, the prognostic value of these imaging signs and clinical features associated with them turn out to be completely different, even in patients with similar extent of lung injury. Haemorrhagic suffusion of lung tissue is often a harbinger of acute respiratory distress syndrome. Some authors believe that haemorrhagic oedema is the cause of exceptionally fast progression of pulmonary abnormalities in patients with an initially small area of lung involvement [21][22][23][24].
On a chest radiograph, thrombosis of small pulmonary vessels is seen as dilation of these vessels in areas of increased density with the ground-glass pattern, but not in consolidation areas, where vessels are not seen on a background of airless lung. This sign is most evident in the cortical lung zones within the first few days (see Figure 1).
The proliferative phase of DAD becomes more prominent over the second or third week of the disease. This period is marked by dissolution of hyaline membranes, formation of immature connective tissue in the alveolar spaces and respiratory bronchioles, hyperplasia of type II alveolar cells, and migration of fibroblasts, monocytes, and macrophages to damaged alveoli. Morphologically, this process is usually viewed as an organising pneumonia. In this phase, the course of the disease depends on the volume and depth of the affected parenchyma and the integrity of the alveolar-capillary membrane [18,20]. When all components are involved and the basal membrane is degraded, which is typically seen in COVID-19, the convalescence phase is almost unavoidably manifests as organizing pneumonia (OP). When lung tissue is not so deeply affected, complete restoration of lung parenchyma is possible without elements of OP, similar to how it happens in common bacterial pneumonia [18,20].
The radiographic findings of OP are well established and frequently described in literature ( Figure 2). These often include signs of cryptogenic OP and OP of known etiologies, such as drug exposure, viral and other infections, radiation damage, and systemic connective tissue diseases [24][25][26][27]. A typical imaging sign of this disorder is the presence of multiple patchy ground-glass density lesions combined with consolidations. They can be located in subpleural, perilobular, or peribronchial regions. An obligatory feature is the air bronchogram sign, i.e. the phenomenon of air-filled bronchi being made visible by the opacification of surrounding tissue. However, the key, but not disease-specific, sing of OP is the reversed halo sign, characterised by a central ground-glass opacity surrounded by a ring of consolidation. It was earlier often referred to as the atoll sign due to its appearance similar to that natural structure. It is typical to visualize on plain radiographs a gradual relocation of opacities from the visceral pleura towards deeper zones of lung parenchyma and apparent lung hyperinflation along the thoracic wall [26]. In some cases, it gives a false impression of emerging cavitating lesions or emphysematous bullae in the cortical regions. So far, no evident temporal distinction has been determined between the DAD phases in the coronavirus infection. In many cases, it is just possible to note the predominance of a particular pathological process at a certain time point. Another specific feature of this infection is a paradoxical prolongation of any phase up to several weeks. Therefore, accurate evaluation of radiographic findings is very meaningful in practice, including in prognosis assessment.
In some patients SARS-CoV-2 specifically affects the immune system, causing uncontrolled and, in most cases, irreversible autoinflammatory response. This is accompanied by a rapid progression of the exudative phase of lung injury [22,28,29]. An X-ray examination reveals a rapid increase in lung involvement coupled with a fast progression of consolidation toward subtotal lung involvement ( Figure 3) [30]. Pleural effusion is a frequent accompanying sign. Clinically, these findings are associated with acute respiratory distress syndrome (ARDS).

Prevalence of the primary radiographic signs
The radiographic signs of lung damage in COVID-19 patients have been described in many studies published in recent months, including large systematic analyses [31,32]. Evaluation of CT data of more than 1,000 patients with verified coronavirus infection revealed the following most typical changes: ground-glass opacities (88.0%), bilateral lung involvement (87.5%), peripheral distribution in lung parenchyma (76.0%), and multilobar involvement (damage to more than one lobe) (78.8%) ( Table 1). Isolated ground-glass opacities or accompanied by consolidation were the most common feature of the disease. Other, less common features, included intralobular and interlobular septal thickening, bronchiectasis, and pleural thickening. Pleural effusion, pericardial effusion, lymphadenopathy, cavitation, the halo sign, and pneumothorax were significantly less common (Table 2).

Changes in the radiographic signs over time
In their study Y.H.Jin et al. [34] described CT findings observed during five stages of COVID-19 infection, which they defined as ultra-early stage, early stage, progression stage, consolidation stage, and dissipation stage: In the ultra-early stage, one or two weeks after contraction, when patients usually had no clinical manifestations, CT scanning showed single or multiple ground-glass opacities, and patchy consolidations or nodules surrounded by ground-glass opacities. In the early stage (early clinical manifestations, 54% in this study), CT revealed single or multiple ground-glass opacities, combined with the crazy paving pattern. In the progression stage (3 -7 days after the onset of clinical symptoms), the disease manifested on CT as large-scale consolidation with air bronchogram inside. In the consolidation stage (the second week of clinical manifestations), CT features may include consolidations in slighter density and smaller range. About two or three weeks after the onset of symptoms, CT scanning may show patchy consolidation, linear opacities, bronchial wall thickening, and interlobular septal thickening.
Following the initial examination, the radiographic features of viral lung damage can evolve in several ways, including the following (Table 3) [3-5, 33, 39-42]: • Reduction in lung involvement accompanied by improvement in pulmonary aeration and dissipation of consolidations into ground-glass opacities and subsequent complete recovery of aerated lung volume. This process is reminiscent of the typical resolution stage of community-acquired pneumonia. • Reduction in lung involvement accompanied by more prominent consolidation and appearance of a typical organising pneumonia pattern. • Stability or even an increase in lung involvement accompanied by more prominent consolidation and appearance of a typical organising pneumonia pattern or separate signs of organising pneumonia. The first variant is the most favourable because it indicates resolution of viral lung damage usually without any residual abnormalities. The second variant indicates a typical reparative process manifested by signs of organising pneumonia that appear in the second phase of DAD. It is a longer process, which may lead to residual pulmonary abnormalities. The long-term prognosis for patients with such abnormalities is now difficult to predict due to little evidence and requires further research. Until these radiographic findings keep improving on follow-up CT scans, they should not be interpreted as "fibrosis" or "sclerosis", i.e. disease resolution with sequelae. As in community-acquired pneumonia, the duration of resolution of lung abnormalities is not limited and does not correlate with the clinical manifestations of the disease.
The third variant of the evolution of radiographic findings may cause difficulty in differentiating positive and negative changes (Figure 4). If the previous examination was performed before the disease peaked, the time when peak lung involvement is reached, a follow-up image may mislead to a conclusion of disease progression. Clinical manifestations are the key to the interpretation of radiographic findings, which can be accurately assessed only in combination with the clinical picture. If respiratory failure does not become more severe, it indicates regression of the disease, even if the changes in radiographic findings seem to be negative. Radiographic signs have some value too. The signs of OP may be suggestive of disease regression. Worsening of ground-glass opacities, consolidations, reticular changes, and, in many cases, pleural effusion without sings of OP more likely suggests negative evolution.

Assessment of the extent of lung involvement
From the prognostic point of view, it is important to assess not only the nature of radiographic findings, but also their extent Assessment of the extent of lung involvement in patients with coronavirus infection can be based on: • imaging data; • data obtained using various semiquantitative scales; • data collected using software for lung density analysis and generation of maps of lung density distribution, including computer-aided design (CAD) and artificial intelligence software solutions. The literature describes a number of similar scales, developed to improve the early diagnosis of the coronavirus infection and assess its severity more accurately (Tables 4, 5). These scales are almost identical in terms of content and differ only in the total score [36,40,42]. It is worth noting that they do not usually consider the type of lung abnormalities (ground-glass opacities and consolidation), except for a 64-point scale developed by F.Feng et al. [44]. Their practical application is, however, hindered, because there is an increasing number of publications suggesting the absence of a direct correlation be-tween the severity of clinical manifestations and the nature and extent of lung injury as assessed by CT [45,46]. It should also be emphasised that such scoring systems are based on the experience of individual researchers, which leads to some inaccuracy of the results when using them in the general population.
Such semiquantitative assessment of the severity (volume and extent) of CT-diagnosed lung abnormalities associated with COVID-19 is possible, but it is not mandatory. These scales can be used by medical facilities as part of an agreed and approved treatment protocol for COVID-19 patients.
A group of Dutch researchers has proposed an assessment scheme for analysing pulmonary findings in patients with suspected COVID-19, using likelihood ratios (CO-RADS), and presented it on the Radiology Assistant website [47] (Table 6). It was developed as an analogue to previously existing scoring systems (PI-RADS, BI-RADS, etc.), which aim to standardise the assessment of likelihood of a certain diagnosis. Since the CO-RADS is not well known to practitioners and not widely used in inpatient facilities, it has no advantages over semiquantitative scales for assessment of lung abnormalities.

Correlation between imaging findings and infection severity
CT does not assess the severity of the disease and can only provide indirect indicators of its prognosis. Although specialists widely use such terms as "severity of CT findings", "severity grading of CT findings", and "severity grades of CT features, such as CT0, CT1, CT2, CT3, CT4, and sometimes CT5", these data primarily reflect the extent of lung involvement. They poorly correlate with the clinical assessment of the severity of the patient's condition at the time of CT examination, including such key parameters as the severity of respiratory failure and the degree of desaturation [45,46].
Multiple publications have reported that patients with severe disease more often have more widespread lung involvement and more frequently develop certain symptoms, for example, reticular changes or pleural effusion. This does not, however, mean that patients with extensive CT or X-ray abnormalities are in critical condition. Importantly, lung lesions can also be found in a considerable number of asymptomatic patients. In the study conducted by S.Inui et al., of 104 cases of coronavirus infection, 76 (73%) were asymptomatic, 41 (54%) of which had pneumonic changes on CT [49]. Our pilot analysis of radiographic and clinical manifestations of 92 patients with COVID-19 showed that two patients (6.9%) with mild infection had extensive lung lesions on a follow CT with > 50% lung involvement, while they did not demonstrate any clinical signs of respiratory failure.

Evolution of changes
In literature, there is no consensus regarding the prognostic value of separate CT findings. P. Lyu et al. showed that the volume of areas with consolidation and the extent of the crazy-paving pattern were significantly greater in severe cases than in mild cases of COVID-19 [50]. The study by F.Liu et al. demonstrated the prognostic value of the extent of ground-glass opacities for predicting the worsening of clinical status. The most common sign reported in this study was a combined pattern (any combinations of ground-glass opacities and consolidation) [51].

Correlation between polymerase chain reaction results and chest computed tomography findings
In most studies, CT findings were generally correlated with PCR results [1,[4][5][6]52]. There have been, however, some publications reporting positive CT findings despite the negative result of the initial screening PCR test. In the study by X.Xie et al. [52], five out of 167 patients with negative initial PCR tests had typical CT features of COVID-19 pneumonia. In all these five patients PRC became positive 2 -8 days after CT examination. In contrast, seven out of the 167 patients had a normal initial CT, despite a positive initial PCR. Within five days after the initial CT, imaging signs of viral pneumonia were observed in one of these patients, while the follow-up data of the other six patients were not reported. Similarly, M.Chung et al. [40] reported that in their study three of 21 patients with a confirmed diagnosis did not have ground-glass opacities or consolidation on their initial CT scans but at the time of re-examination two of them demonstrated positive CT findings. No information was reported about the results of their confirmatory laboratory tests.
H.Kim et al. [53] performed a meta-analysis to evaluate diagnostic values of chest CT and PCR. For chest CT scans, the positive predictive value (PPV) ranged from 1.5 to 30.7%, and the negative predictive value (NPV) ranged from 95.4 to 99.8%. For PCR, the PPV ranged from 47.3 to 96.4%, whereas the NPV ranged from 96.8 to 99.9%. The authors reported that the pooled sensitivity was 94% for chest CT and 89% PCR. The pooled specificity for chest CT was only 37% (95% CI: 26, 50%). The authors suggested that considering a low specificity of CT there is a large gap between PPV levels of chest CT and PCR in low-prevalence regions, especially in areas with a prevalence less than 10%.
These data suggest a high probability of false positive CT findings. This obviously dictates the need for fol- Table 4 Lung Severity Score scale [42] Таблица 4 Шкала степени поражения легких [42] Degree of lung involvement, % Score 0 0  Interpretation: The total score of 19.5 or above was predictive of a severe course with a 83.3% sensitivity and a 94% specificity

Imaging and etiological diagnosis
Imagining modalities have a high sensitivity for lung involvement in cases suspected for COVID-19. Nevertheless, all of them, including CT, show a low specificity because the same findings can be observed in different lung infections. Certain radiographic findings and their combinations (Table 2) suggest the presence of COVID-19 with a certain probability [33]. The differential diagnosis should primarily include other types of viral pneumonia caused by other coronaviruses (SARS and MERS) and adenoviruses [2,54,55]. Pneumonia caused by influenza viruses, adenovi ruses, and pneumoviruses (respiratory syncytial virus and human metapneumovirus) may have CT features similar to those of COVID-19. However, they rarely manifest by subpleural lesions. In addition, due to certain pathogenetic factors ground-glass opacities or consolidations can develop in the centrilobular regions [55].

Conclusion
Cardiogenic oedema is another complication of COVID-19. Its manifestation can be associated both with a direct cytotoxic effect of the virus on the myocardium and vascular endothelium and exacerbation of chronic heart failure [57][58][59]. CT findings include bilateral subtotal consolidations and ground-glass opacities with air bronchogram located in the middle and upper lung zones and intralobular, interlobular, and peribronchovascular interstitial thickening. There are also gradients in lung density, depending on the patient's position (supine, prone) and hyperinflation of basal segments. As circulatory failure progresses, pulmonary abnormalities also become more prominent, patients develop plural effusion (unilateral or bilateral), and enlargement of the left heart becomes visible by imaging modalities [33,60].