The main finding of our study is that LUSS at discharge in survived patients was still elevated and not significantly different from the score observed on the day of admission to ICU. The most represented LUS pattern in the total collected videos on admission was type C. All 28 patients showed non-translobar consolidations, mostly characterized by a mantellary distribution in posterior and lateral zones. These peripheral consolidations may be considered the hallmark of the COVID-19 pneumonia, together with the other signs of the interstitial acute damage. Similar consolidations are observed also in lung infarcts due to pulmonary embolism (PE), early stages of bacterial pneumonia, small subpleural lung abscess, metastasis and subpleural foci of lung cancer . Thus, the differential diagnosis of the origin of these consolidations remains to be explored as it has been demonstrated that COVID-19 may be complicated by PE and bacterial cross infections .
However, in our cohort, procalcitonin values suggested a low probability of concomitant bacterial superinfection at ICU admission . We believe LUS cannot differentiate if those consolidations are part of a patchy and multiform ongoing interstitial viral infection or if they are due to infarction in segmental or subsegmental embolism. Therefore, multiorgan integration of LUS with venous and cardiac ultrasound may be of help .
Peng et al.  reported a variety of patterns, including multifocal small, non-translobar, and translobar consolidations with occasional mobile air bronchograms. The same authors did not detail the prevalence of translobar consolidations with air bronchograms. We speculate that the prevalence of this LUS pattern could be influenced by the timing of LUS acquisition during the time course of the COVID-19 pneumonia, by its clinical severity and probability of superinfections. In our study only one patient, with a previous history of chronic renal failure, showed both peripheral consolidations and bilateral basal translobar consolidations with air bronchograms and bilateral large pleural effusion with elevated values of pro-BNP and procalcitonin on admission.
On the day of admission, the A pattern associated with normally aerated lung was mostly present in the anterior and superolateral lung regions. The more represented B pattern was the B2, in most cases observed by moving the probe from the anterior to the lateral and posterior areas in a quarter of the videos recorded at admission. The B-lines patterns were often associated with small subpleural subcentimetric echo-poor areas, had a dishomogeneous patchy distribution and were evident despite positive end-expiratory pressure (PEEP), which was given at a median value of 12 [10–12] cmH2O. These findings suggest a diffuse peripheral parenchymal lung damage that is not responsive to PEEP during the initial course of the disease in severe hARF and is unrelated to cardiogenic alveolar edema .
Similarly to admission, LUS performed in our patients at discharge was characterized by normal A pattern, mostly represented in the anterosuperior regions, and B pattern in more than one fourth of the acquired videos, mostly localized in the lateral and posterior zones, with higher prevalence of B2 pattern over B1. At discharge our patients had the greatest prevalence of C pattern, which was present in 61.3% of the total 336 collected videos (206/336), regardless of the clinical outcome (death or improvement).
It is noticeable that a difference in LUSS at admission and discharge was not detected, even in survived patients, suggesting that variation of LUSS, which has been validated by Soummer et al.  for patients affected by other conditions originating severe hARF, could not be a useful tool in COVID-19 patients. Other recent studies6 seem to demonstrate that the LUSS severity correlates with CT scan and the clinical picture. However, how chest imaging may be related to the evolution of the disease remains still not proven and the trend observed in our patients seems to exclude a role for the traditional LUSS in COVID-19. Indeed, many evidences suggest that the clinical characteristics of COVID-19 related ARDS are different from classical ARDS observed before this pandemic . This could be the reason of our findings: in ARDS caused by other conditions than COVID-19, aeration loss predominates in dependent lung regions (focal ARDS), with parts of upper lobes remaining more aerated . In COVID-19 patients, the presence of non-translobar, mantellary, widespread consolidations leads to an increase in median LUSS and therefore makes the already described LUSS thresholds hardly applicable to guide clinical decisions . Some authors  suggest a modified method to calculate LUSS, in order to overcome some limitations of the traditional LUSS and to better quantify the observed findings. Anyway, we believe that the use of the different scoring systems could afflict the absolute value of LUSS but not the difference between the two LUSS values found at admission and discharge.
In the videos collected both at admittance and discharge, lung sliding was always present. In agreement with other authors , pleural effusions were uncommon and detected only in one patient who also showed translobar consolidations. Their presence could be due to heart failure or concomitant bacterial pneumonia.
Our study has some limitations. First, the limited sample size does not allow generalization of our results but can only be useful to fix a hypothesis that needs to be confirmed in a large multicentric population study. Moreover, our examination was not optimal as the LUS videos of the posterior paravertebral regions were partially hampered by the supine decubitus in our sedated and mechanically ventilated patients. We believe that in ICU setting, the optimal examination of COVID-19 patients should always include the whole paravertebral areas from the superior to the inferior chest area. Of course, this requires the collaboration of another member of the staff who will be of help in maintaining the patient in lateral decubitus during the posterior examination.