Results showing how well logistic regression classifiers can identify which sub-dataset a CXR is from within the COVIDx dataset, and how well classifiers can identify which dataset a “COVID-19+” CXR is from across both the COVIDx and CC-CCII datasets.

We report AUC values as averages of the one-vs-all binary AUCs between all classes, and accuracy (ACC) as the average accuracy over all classes. We observe that the representations generated by the classifiers, even from masked/equalized inputs, contain enough information to accurately identify the sources of the imagery in both cases.