Back

Prognostic value of visual quantification of lesion severity at initial chest CT in confirmed Covid-19 infection: a retrospective analysis on 216 patients

Taieb, E.; Labani, A.; Ruch, Y.; Danion, F.; Oberlin, M.; Bilbault, P.; Leyendecker, P.; Roy, C.; Ohana, M.

2020-05-30 radiology and imaging
10.1101/2020.05.28.20115584 medRxiv
Show abstract

Rationale and ObjectivesStudies suggest an association between chest CT findings assessed with semi-quantitative CT score and gravity of Covid-19. The objective of this work is to analyze potential correlation between visual quantification of lesion severity at initial chest CT and clinical outcome in confirmed Covid-19 patients. Materials and MethodsFrom March 5th to March 21st, 2020, all consecutive patients that underwent chest CT for clinical suspicion of Covid-19 at a single tertiary center were retrospectively evaluated for inclusion. Patients with lung parenchyma lesions compatible with Covid-19 and positive RT-PCR were included. Global extensiveness of abnormal lung parenchyma was visually estimated and classified independently by 2 readers, following the European Society of Thoracic Imaging Guidelines. Death and/or mechanical ventilation within 30 days following the initial chest CT was chosen as the primary hard endpoint. Results216 patients (124 men, 62 years-old {+/-} 16.5, range 22 - 94 yo) corresponding to 216 chest CT were included. Correlation between lesion severity and percentage of patients that met the primary endpoint was high, with a coefficient {rho} of 0.87 (p = 0.05). A greater than 25% involvement was significantly associated with a higher risk of mechanical ventilation or death at 30 days, with a Risk Ratio of 5.00 (95% CI [3.59-6.99]). ConclusionThis retrospective cohort confirms a correlation between visual evaluation of lesions severity at initial chest CT and the 30 days clinical outcome of Covid-19 patients and suggests using a threshold of greater than 25% involvement to identify patients at risk.

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.