Figure 6
From: Assessing clinical applicability of COVID-19 detection in chest radiography with deep learning
Equipment-wise evaluation of \(M_{MAnnot}\) on the CHVNGE dataset. Top left: model performance on all CXRs. Top right: model and radiologist performance on CXRs annotated by radiologists (Carestream not shown due to limited data). Average model performance for all folds is shown as a line and shaded region corresponds to the 95% confidence interval. Radiologists’ performance is shown considering as positives only CXRs marked as Indicative of COVID-19 (\(\bullet \)) or Indicative of COVID-19 and Undetermined (\(\blacklozenge \)).
