Ces radiol. 2020, 74(3):180-188 | DOI: 10.55095/CesRadiol2020/028

The assessment of the lung involvement in COVID-19 pneumonia using automatical analysis employing the artificial intelligence algorithmOriginal article

Martin Vítovec¹, Jan Baxa¹, Hynek Mírka¹, Thomas Flohr², Bernhard Schmidt², Jiří Ferda¹

¹ Klinika zobrazovacích metod LF UK a FN, Plzeň

² CT research division, Siemens Heathineers, Forcheim

Aim: To compare the extent of the lung involvement hospitalized at the non-intensive care unit and intensive care unit using the analysis assisted by the artificial intelligence.

Method: The retrospective analysis of the sample of 50 patients, who underwent chest CT indicated due to the suspected SARS-CoV-2 infection and followed by the confirmation of the diagnosis by the PCR test. The patients were investigated during spring wave of the epidemic between March and April 2020. The sample was split into two groups hospitalized at the non-ICU (31 pt.) and ICU (19 pt.). We analyzed the data in these patients using software prototype Interactive CT Pneumonia Analysis provided by Siemens Heathineers. The percentage of the consolidation and ground glass opacities were quantified in lobes and lungs.

Results: The mean percentage of the opacities related to the whole volume of lungs in patients hospitalized in ICU was 31.56%, 11.25 in patients in non-ICU respectively; the consolidation was affected 5.2% in ICU group, 2.09% in non-ICU group. Domination of right lower lobe was in 39.17% in ICU patients, 18.65% in group of non-ICU patients.

Conclusion: The assessment of the lung parenchyma load using lung infection AS assisted analysis enabling the quantification of the affected volume, makes possible to score the patients and making the parametric staging, it allows to estimate the need of intensive care.

Keywords: COVID-19 pneumonia, SARS-CoV2 infection, chest CT, infection load, artificial intelligence

Grants and funding:

Podpořeno projektem MZ ČR Koncepční rozvoj výzkumné instituce 00669806 - FN Plzeň.

Accepted: September 15, 2020; Published: September 1, 2020  Show citation