Examples showing annotations of the radiologists and the simultaneous truth and performance level estimation (STAPLE)-generated consensus ROI. (a,e) show the CXR image instances from the Shenzhen TB CXR-Subset-2 and Montgomery TB CXR datasets, respectively; (b,c) show the annotations of R2 and R3 (bounding boxes in red and green), respectively, for the given CXR instance from the Shenzhen TB CXR-Subset-2 dataset; (f,g) show the annotations of R2 and R1 (bounding boxes in red and yellow), respectively for the given CXR instance from the Montgomery TB CXR dataset; (d,h) show the STAPLE-generated consensus ROI (bounding boxes in blue) from the experts’ annotations for the given CXR instance from the Shenzhen TB CXR-Subset-2 dataset and Montgomery TB CXR dataset, respectively.