Lung injury in patients with or suspected COVID-19 : a comparison between lung ultrasound and chest CT-scanner severity assessments, an observational study
Benchoufi, M.; Bokobza, J.; Anthony Chauvin, A.; Dion, E.; Baranne, M.-L.; Levan, F.; Gautier, M.; Cantin, D.; d Humieres, T.; Gil-Jardine, C.; Benenati, S.; Oberlin, M.; Martinez, M.; Kahn, N.; Vicaut, E.; Bourrier, P.; Diallo, A.
Show abstract
BackgroundChest CT (CT) is the reference for assessing pulmonary injury in suspected or diagnosed COVID-19 with signs of clinical severity. We explored the role of lung ultrasonography (LU) in quickly assessing lung status in these patients. MethodseChoVid is a multicentric study based on routinely collected data, conducted in 3 emergency units of Assistance Publique des Hopitaux de Paris (APHP); 107 patients were included between March 19, 2020 and April 01, 2020 and underwent LU, a short clinical assessment by 2 emergency physicians blinded to each others and a CT. LU consisted of scoring lesions in 8 chest zones from 0 to 3, defining a severity global score (GS) ranging from 0 to 24. CT severity score ranged from 0 to 3 according to the extent of interstitial pneumonia signs. 48 patients underwent LU by both an expert and a newly trained physician. FindingsThe GS showed good performance to predict CT severity assessment of COVID-19 as normal versus pathologic: AUC=0.93, maximal Youden index 1 with 95% sensitivity, and 83% specificity. Similar performance was found for CT assessment as normal or minimal versus moderate or severe (n=90): AUC 0.89, maximal Youden index 7 with 86% sensitivity, and 78% specificity. Good agreement was found for zone scoring assessed by new trainee (30mn theory + 30mn practice) and expert (n=14,14*8 checkpoints), weighted kappa 0.85-1; moderate agreement was found for new trainee (n=48, 30mn theory) and expert, kappa 0.62-0.81. InterpretationGS score is a simple tool to assess lung damage severity in patients with suspected or diagnosed COVID-19. Comparing the performance of new trainees and expert physicians opens a path for adoption beyond the scope of experts. LU is a good candidate for patients triage, especially in case of CT availability issues.
Matching journals
The top 4 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- Quantitative analysis of chest computed tomography of COVID-19 pneumonia using a software widely used in Japan 94%
- Accuracy of deep learning based computed tomography diagnostic system of COVID-19: a consecutive sampling external validation cohort study 93%
- Organizing Pneumonia of COVID-19: Time-dependent Evolution and Outcome in CT Findings 92%
Similar papers in this journal
Similar papers in this journal
- Intensive care risk estimation in COVID-19 pneumonia based on clinical and imaging parameters: experiences from the Munich cohort 92%
- Screening for Right Ventricular Dysfunction in the Emergency Department Using a Smartphone ECG Analysis Application: An External Validation Study with Acute Pulmonary Embolism Patients 91%
- Artificial Intelligence in laryngeal endoscopy: Systematic Review and Meta-Analysis 90%
Similar papers in this journal
- The incremental value of computed tomography of COVID-19 pneumonia in predicting ICU admission 93%
- High-Dimensional Multinomial Multiclass Severity Scoring of COVID-19 Pneumonia Using CT Radiomics Features and Machine Learning Algorithms 92%
- Content-based image retrieval assists radiologists in diagnosing eye and orbital mass lesions in MRI 91%
Similar papers in this journal
- Passive Microwave Radiometry (MWR) for diagnostics of COVID-19 lung complications in Kyrgyzstan 93%
- Auto-detection of motion artifacts on CT pulmonary angiograms with a physician-trained AI algorithm 92%
- Volumetric lung cancer screening reduces unnecessary low-dose computed tomography scans: results from a single-centre prospective trial on 4,119 subjects 91%
"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.