TOMOGRAPHIC FINDINGS IN COVID-19 LUNGS: THE CONTRIBUTION OF ARTIFICIAL INTELLIGENCE

Le résumé fourni par la source

Since the beginning of the Coronavirus Disease (COVID-19) pandemic, chest computed tomography has been a strong ally in the diagnosis and monitoring of patients. New tools have been developed to make lung analysis more objective, such as 3D Slicer and its Lung Analyzer tool. This study aimed to describe the tomographic findings found in patients with COVID-19 admitted to the Intensive Care Unit using this tool. Method. One hundred and one sets of tomographic images of patients admitted between March 2020 and December 2021 with a positive diagnosis of COVID-19 were randomly selected. The Lung Analyzer was used to perform the analysis, quantifying the “emphysematous”, “infiltrated”, “collapsed” and “totally affected” lung in milliliters and their percentage. Results. The findings of emphysema remained around 20% of the volume. Infiltrates, collapsed lungs and the percentage of total involvement did not show this similarity, with a significant plurality of results. In pulmonary infiltrates, an average of 32.98% of involvement was obtained, with a standard deviation of 0.095. For the percentage of collapsed lung, we have an average of 11.48% and a standard deviation of 0.067. The total percentage of affected lung was 44.3% on average, with a standard deviation of 0.15. Conclusion: The description of findings found by the software can be valuable for the identification and quantification of lung lesions in patients with COVID-19, reducing the subjectivity of reports and helping to better understand lesions, sometimes not so visible to the human eye. Its use does not exclude the need for experienced radiologists for the best evaluation.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.