All three CAD technologies showed limitations in handling erroneous images, with different strengths and weaknesses. CAD1 was most effective at detecting errors but generated more false alarms with standard images. CAD3 processed almost all images regardless of quality, largely failing to flag technical errors.

CAD1 and CAD2 demonstrated good accuracy in detecting specific non-TB pulmonary findings in adults and children. These results highlight the potential of CAD technologies to support comprehensive CXR interpretation beyond TB screening, particularly for identifying non-TB abnormalities.