Automatic vs. visual analysis for lung aeration quantification with lung ultrasound in COVID-19 ARDS

ABSTRACT Qualitative lung ultrasound is a useful tool for the assessment of acute respiratory failure to facilitate differential diagnosis; during the COVID-19 pandemic, it has been largely used to early identify lung involvement of the disease. A quantitative approach is also possible, allowing to quantify the lung aeration and density on the basis of the visualized artifacts in 12 standard thoracic regions ? i.e. computation of the lung ultrasound score. The visual analysis performed by an expert operator showed strong correlation with lung density as assessed by quantitative CT scan, the gold standard technique. The quantitative approach allows using lung ultrasound not only as a diagnostic, but also as a monitoring tool; it has already been applied to monitor the response to antibiotics in ventilator-associated pneumonia, the recovery of lung disease in patients supported by ECMO, the indication to weaning from mechanical ventilation and the response to PEEP titration or pronation in ARDS. The main limitation of lung ultrasound is the need of adequate training and expertise, mainly for an advanced skill as quantitative analysis. Automation may help in overcoming this limitation, allowing a wider spread of the technique. Different automatic analyses for pleural assessment have been proposed so far. Aim of the present study is to compare the visual analysis performed by an expert operator ? considered as the reference technique ? to an automatic analysis performed by a new software developed by Lecce CNR for the computation of the lung ultrasound score and the assessment of lung aeration in patients affected by COVID-19 pneumonia and admitted to Pavia ICU. The analysis was performed on a pilot population of 21 patients, corresponding therefore to 252 scans.