Sensitivity of Lung US compared to Chest CT for the screening of COVID-19: preliminary report of our experience


 As lung ultrasound (LUS) is a noninvasive, radiation-free, repeatable and portable imaging tool suitable for a point-of-care use, several recent literature reports have emphasized its role as the ideal screening tool for SARS-CoV2 pneumonia. To evaluate the actual diagnostic accuracy of LUS for this purpose, we performed a systematic comparative study between LUS and CT scan findings in a population of 82 patients hospitalized because of COVID-19. LUS and Chest CT have been performed in all patients within 6-12 hours from the admission. The sensitivity of LUS in assessing typical CT findings was 60%. Despite LUS detected consolidations adherent to pleural surface in all cases, it was not able to detect all the consolidations assessed at CT scan (p=0.002), showing a risk to underestimate the actual disease’s extent. Moreover, only 70% of pleural surface is visible by LUS. Considering that the specificity and the positive predictive value of the same LUS signs may be lowered in a normal setting of non epidemic COVID-19 and in case of pre-existing cardio-pulmonary diseases, LUS use should not be indicated for diagnosis of COVID-19. However, it may be very useful for the assessment of pleural effusion and to guide safer fluid drainage.


Introduction
Since the initial cluster of pneumonia cases in Wuhan in December 2019 1 , the so-called SARS-CoV-2 has sudden spread globally and the World Health Organization (WHO) declared the novel coronavirus disease  as "a pandemic" on 11 March 2020 2 .
The clinical presentation of COVID-19 varies remarkably from one patient to another, going from asymptomatic forms to life-threatening acute respiratory distress syndrome requiring admission to intensive-care-units (ICUs). The diagnosis may be challenging. Chest X-Ray generally represents the faster, widely available rst-line imaging method for symptomatic patients in the Emergency Departments. Computed Tomography (CT) has a higher sensitivity than X-ray in detecting ground glass opacities (GGOs) in the early phase of the disease 3,4,5 . However, its use for the rst assessment of COVID-19 infection in epidemic areas is not an easy issue. Indeed, it implies a huge burden on Radiology Departments, the designation of dedicated CT machines and the application of severe infection control procedures 6,7 .
Lung ultrasound (LUS) is a safe, noninvasive, radiation-free, repeatable, cost-effective and well-tolerated imaging tool. Moreover, it is portable and usable as immediate point-of-care method. LUS has the ability to detect any process involving subpleural pulmonary areas, if adherent to the accessible pleural surface 8 . However, about the 30% of pleural surface and the deeper lung parenchyma are not accessible to LUS due to technical limitations 9 The typical Chest-CT ndings of COVID-19 include bilateral subpleural and lower lobe located GGO opacities and/or consolidation 5 , that are likely visible to LUS 10 . Frequently reported LUS ndings are the appearance of multiple, con uent or not, B-lines, a thickened and irregular pleural line and small super cial or large areas of consolidation 11,12,13 . Also minimal or rarely larger pleural space uid effusions has been evidenced 14,15 .
With COVID-19 pandemic outbreak, several literature reports have emphasized the role of LUS as an useful screening tool for Emergency Department diagnosis, pre-hospitalization triage, ICU decisions (regarding ventilator need and weaning), and treatment monitoring 10,13 . However, these works included few patients, did not indicate the equipment setting and did not consider the main comorbidities of COVID-19 patients, thus reaching premature and fascinating conclusions. In addition, despite typical LUS ndings may show good sensitivity and positive predictive values in the context of COVID-19 epidemic (i.e. high "a priori" probability of disease in the presence of respiratory symptoms), the ability of LUS to rule out COVID-19 in normal condition is far from su cient, as the same US patterns overlap with several other pleuro-pulmonary conditions 16 . At present, there is still a lack of data to establish the appropriate use of LUS in the diagnostic workup of COVID-19. A deeper knowledge of LUS sensitivity to different stages of pulmonary involvement is needed.
With this background, we performed a systematic comparison between LUS ndings and the respective CT scan appearances in a COVID-19 cohort with the aim to estimate the actual diagnostic accuracy of LUS in detecting COVID-19 related abnormalities in case of typical or not CT lesions.

Participants And Methods
Participants. We performed a prospective comparative study between LUS and CT ndings in a cohort of COVID-19 patients. We analyzed data of the rst consecutive 82 patients hospitalized in our referral COVID-19 center in Southern Italy, "Fondazione Casa Sollievo della Sofferenza" Research Institute, located in San Giovanni Rotondo, Foggia, Italy, from March 19 and April 13, 2020.
All the evaluated patients had a con rmed diagnosis of COVID-19 by a positive RT-PCR assay for SARS-CoV-2 on nasopharyngeal swabs collected at admission. Among the 82 inpatients enrolled, 59 patients presented with persistent fever, cough and fatigue and were admitted in a COVID-19 Emergency Department, 10 presented with acute dyspnea and pneumonia and were hospitalized in a COVID-19 ward and 13 patients presented with a severe acute respiratory distress syndrome and were admitted in a COVID-19 intensive or sub-intensive care units. COVID-19 severity was assessed at admission on the basis of a series of measurable laboratory and clinical parameters, including blood C-reactive protein (CRP) values and peripheral oxygen saturation (SpO 2 ). A Chest-CT and lung US examination was performed in all patients concurrently, i.e. within 6-12 hours from the admission to the hospital.
Ethical approval and informed consent. The study was approved by the institutional Ethics Committee of "Fondazione Casa Sollievo della Sofferenza" Hospital and was carried out according to the principles of the Declaration of Helsinki. All the participants or their legal guardians provided informed written consent for all the procedures. There is no identifying information or image in the article.
Chest-CT. Chest-CT examinations was performed using a multi-detector CT scanner with 64 channels. The detailed parameters for CT acquisition were as follows: tube voltage, 120 kVp; tube current, standard (reference mAs, 60-120); slice thickness, 0.5 mm; reconstruction interval, 0.3-1.0 mm. All CT images were acquired at full inspiration (impossible in few severely ill patients) with the patient in the supine position and without contrast medium. Chest CT scans were interpreted by a radiologist, with 32 years of experience in thoracic imaging. All the CT examinations were reviewed by a second expert in thoracic imaging to reach a consensus.
Classi cation of patients according to Chest-CT ndings. Given the variable CT appearance in COVID-19, the RSNA Expert Consensus Statement on Reporting Chest-CT Findings Related to COVID-19 17 has recently suggested an useful classi cation for reporting CT ndings potentially attributable to COVID-19.
We have classi ed Chest-CT results in our COVID-19 population according the four categories proposed by Simpson et al 17 : 1. Typical CT appearance: GGOs, showing a round morphology or a "crazy paving" pattern, with or without consolidations, in a peripheral, posterior, and diffuse or lower lung zone distribution; 2. Indeterminate CT appearance: focal or diffuse GGOs without a clear distribution; 3. Atypical CT appearance: lobar or segmental consolidations, cavitations, tree-in-bud opacities with centrilobular nodules and all those alterations that are reported to be uncommon or not occurring in COVID-19 pneumonia and are more typical of other diseases; 4. Normal CT appearance: absence of parenchymal abnormalities attributable to infection.
Findings of those four categories were then systematically compared them with the corresponding US ndings.
Lung US. Lung US examination was performed with an Esaote MyLab-25 GOLD and My-Lab Twice (Esaote-Biomedica, Genoa, Italy) using a multifrequency convex probe (3-5 MHz and 3-8 MHz) and an adequate setting for the adult thoracic study (gain: max 50%, focus pointed on the hyperechoic pleural line, activation of the tissue harmonic).
The chest was examined in each patients by exploring lung elds from the bases up to the ipsilateral apexes in the following regions: 1) Anterior: along the parasternal, mid-clavicular and anterior-axillary lines; 2) Lateral: along the mid-axillary and posterior-axillary lines; 3) Posterior: along the mid-scapular and para-vertebral lines. The approximate duration of the entire lung US examination was 15 minutes.
The lung US examinations were performed and interpreted by 3 expert sonographers, with 10-32 years of experience in diagnostic and interventional ultrasound. The Chest-CT scan of each patient was blinded to the sonographers during the exam. Examination videoclips were recorded and later blindly re-examined by another expert sonographer with 20 years of experience in lung ultrasound in order to asses intra-and inter-operator variability. Lung US examination was focused on the detection of the following ndings: In only 1 patient, a focal GGO reaching the pleural surface and measuring 20 mm in diameter at Chest-CT was associated with a non-speci c smooth subpleural nodulation at US, measuring approximately 9 mm.
The Chest-CT crazy pattern of bilateral, patchy or extensive, peripheral GGOs associated with smooth interlobular and intralobular septal thickening was identi ed at LUS in a total 30/58 patients (52%), with an echographic pattern consisting in a blurred and thickened hyperechoic pleural line, with or without associated subpleural hypoechoic lung striae, and B-lines below the pleural line ( gure 1).
LUS was able to identify bilateral pulmonary consolidations, with predominately ill-de ned margins and Related to COVID-19 17 , with the corresponding US ndings in our COVID-19 patients (

Discussion
Generally, a screening test should be highly sensitive, allowing a large proportion of subjects affected by the disease to be correctly classi ed as having such disease. In our case series, Chest-CT identi ed COVID-19 lung abnormalities in all patients. On the contrary, the detection of COVID-19 lung abnormalities by LUS was impossible or inadequate in more than one third of cases. Therefore, LUS showed a lower global sensitivity (of about 52% in our casuistry) in detecting pulmonary lesions, compared to Chest-CT.
Indeed, with respect to CT-scan, LUS shows several technical limitations in lung parenchyma assessment.
In particular, less than 70% of the pleura-pulmonary surface is visible with ultrasound 8,18 : part of the chest is US-probe-blinded, due to the overlying bone structures (i.e. ribs, scapulae); part of the lung is not adherent to pleural surface, preventing full US visibility of also big lesions due to the interposition of lung air content.
Intuitively, typical Chest-CT ndings are more likely visible to LUS, being them distributed in a peripheral and posterior, diffuse or lower lung, zone. However, also the diagnostic accuracy of LUS in assessing typical CT ndings in our case series was 65% (i.e. lower than that of Chest-CT), with a percent of true positive subjects (i.e. a sensitivity) of 60% and a probability of not having any typical alterations at CT scan in case of negative LUS (i.e. a negative predictive value) of only 28%. As a further con rm, the Cohen's kappa coe cient between Chest-CT and Lung US in diagnosing Typical COVID-19 lung lesion was weak.
In particular, the sensitivity of an echographic pattern consisting in a blurred and thickened hyperechoic pleural line with B-lines below it in detecting a CT peripheral "crazy-paving" pattern of GGOs was of only 52%. Nevertheless, this artifactual US pattern is an unreliable, poor reproducible and potentially misleading diagnostic approach, due to variability among different intra-and inter-operator B-lines counting and ultrasound scan settings 19 . Indeed, the erroneous use of a medium-to-low frequency probe or excessive total gain and the lack of tissue harmonic imaging can generate a large number of such artifacts. Furthermore, the increase in the pleural line movement rate in dyspneic patients or the simple change of positioning of the probe with respect to the curvature of the patient's chest can in uence the perceptual semi-quantitative evaluation of B-lines 19 .
Similar to chest-CT, LUS was able to assess subpleural consolidations in all cases (12/12). Despite this, it was possible to assess the exact number and extension of such consolidations only in a third of them. The reason for this discrepancy relies on the possibility that consolidated lung may be obscured also by a very thin layer of air in the more super cial lung or may be located in parts of the lung that are not accessible on US, such as the retro-scapular region, the mediastinal surface area or the costo-vertebral junction region. In such cases, the risk is to miss the detection of some lesions and/or underestimate the actual disease's extent ( gure 2).
The speci city and the positive predictive value in assessing US alterations in presence of typical CT ndings were 92% and 97% respectively in our case series. This suggests that the probability of having typical CT patterns at Chest CT in a subject with positive LUS is high as well as the proportion of subjects without typical CT lesions and a negative LUS. This data, however, should be handled with care. In fact, this does not mean that LUS is highly speci c in COVID-19 pneumonia. If correctly performed, LUS has a high speci city and a very high positive predictive value in assessing lesions that are adjacent to the viewable pleural surface, but, once again, it may not allow to visualize any alterations, if these are located in areas not accessible to US, such as in the case of the indeterminate Chest-CT lesions in our patients.
Nevertheless, it should be considered that the speci city and the positive predictive of the same LUS signs may be lowered in a normal setting of non epidemic COVID-19. Indeed, the sonographic ndings described in our case series and in the current available literature on this topic 11,12,13,14,15 shows a considerable overlap with many other lung diseases. An irregular pleural line with increased B-lines may be visible in ARDS, heart failure, nephrotic syndrome, bacterial pneumonia, other viral pneumonia, also minimal pleural effusion, hydropneumothorax, brosis, pulmonary contusion, exacerbations of chronic obstructive pulmonary diseases and neoplastic lymphangitis 16 . Subpleural consolidations may be visible in other viral pneumonia, non viral pneumonia, atelectasis and lung cancer 16 and their LUS patternconsisting in mixed hypo-echogenicity, with irregular, scarcely de ned borders -is non-speci c, not allows to distinguish one condition from another. Furthermore, some of these overlapping conditions may even be pre-existing in COVID-19 patients (especially in more severe cases) and LUS is often unable to discern a COVID-19 diagnosis in a population with such pre-existing cardiothoracic conditions, including chronic obstructive pulmonary disease, interstitial lung disease, cardiovascular disease and malignancies with cardiothoracic involvement 18,20 .
The American College of Radiology (ACR) does not recommend the use of chest CT to screen patients for COVID-19 pneumonia, due to the high possibility of typical CT ndings' overlapping with other viral and non-viral conditions 21 . Indeed, other preexisting pathologies may resemble the atypical or rare CT manifestations of this viral pneumonia 22 . Moreover, also con rmed positive patients can show negative chest CT 23 . For these reasons, it has been recommended a sparing use of CT, that has to be reserved for hospitalized, symptomatic patients with speci c clinical indications. Viral testing remains the only speci c method of diagnosis, whose con rmation is required, even if radiologic ndings are suggestive of COVID-19.
In our COVID-19 patients, LUS resulted falsely negative in most cases, showing much less sensitivity than Chest CT in assessing disease-related lesions. This inevitable result is conform with the physical characteristics and limitations of the pleuro-pulmonary ultrasound examination and highlights LUS inadequacy for a screening purpose 24 . Furthermore, due to the high non-speci city of US ndings and the di culty to discriminate possible pre-existing cardio-pulmonary comorbidities, the incidental detection of alterations that could be attributable to COVID-19 pneumonia should be regarded and classi ed with much more attention. To date, no comparative studies have been performed between COVID-19 patients and patients with other possible overlapping cardio-pulmonary diseases. Therefore, LUS is not suitable for formulating diagnostic hypotheses based on conjecturally speci c clues that are, actually, not reproducible, di cult to demonstrate, confusing in case of pre-existing comorbidities and have never received unanimous consensus or solid support 24,25 . Moreover, one must ask how a method like US, which only visualizes a small part of the pulmonary parenchyma could ever enable a reliable assessment of diseases extent and severity 16,26 . With these consideration in mind, if the role of CT in COVID-19 is still debated, that of LUS is not unexpectedly even more uncertain.
Otherwise, LUS is an excellent imaging method for the study of pleural effusions, being able to detect also minimal amount of liquid. In this respect, LUS is greatly superior to other standard thoracic imaging techniques (i.e. both chest X-Ray and CT), as showed also in our report. Although pleural effusions are not a typical feature of COVID-19, ultrasound may therefore be very useful to guide pleural punctures for safer uid drainage and for the assessment of the changes in the amount of pleural uid over time. It would be interesting to assess, by further studies on large case series, if LUS could prove useful in the follow-up during therapy of consolidations ascertained by CT and visible by ultrasound (i.e. consolidation strictly adherent to the super cial pleura).
In conclusion, to date, with the exception for US-guided procedures and interventions, the use of LUS should not be indicated for diagnostic screening and monitoring of COVID-19 patients. As bedside US implies a prolonged exposure of operators to patients and vice versa (longer and closer than that of a CT examinations), any US examination in patients with COVID-19 should be limited to essential imaging procedures for the ongoing clinical management of the patient (not just lung US, but rather thyroid, carotid artery, liver, renal or any other examinations), with operators protected by all the necessary personal protective equipment (PPE) in order to avoid infection transmission. Other LUS uses in patients with COVID-19 should be justi ed only within the context of a controlled research study.

Declarations
This study involved human subjects.
The author con rmed that all appropriate ethical guidelines for the use of human subjects have been followed, any necessary IRB and/or ethics committee review has been obtained, and information about the IRB/ethics committee is included in the manuscript.
The author has con rmed that all necessary patient/participant consent or assent has been obtained and the appropriate institutional forms have been archived. If the IRB/ethics committee waived the requirement for patient/participant consent or assent, an explanation for the waiver is included in the text.
The author has con rmed that a statement listing potential con icts of interest or lack thereof is included in the text.

Author contributions
All the authors contributed to the conception and design of the study. Q.C.M.I., M.A., D.L., M.V. and M.S. contributed to acquisition and interpretation of data, to drafting the work, to write the main text of the manuscript and revising it critically. R.R., M.M.M., R.G., S.A. and F.B. contributed to acquisition and interpretation of data, including lung ultrasound and Chest CT. All authors reviewed the manuscript and approved its submitted form.

Competing interests
All the authors certify to have not any actual or potential con ict of interest to disclose, including any nancial, personal or other relationships with other people or organizations that could inappropriately in uence or bias the work.