The current FDA-approved second-reader usage of CAD has set a low bar for CAD to be considered a success
The radiologist interprets a case without CAD and then the CAD prompts are shown and the radiologist has the opportunity to revise their interpretation. Performance is measured without and with CAD, and if the difference is not significant, CAD is considered a success. This is an unsatisfactory approach to assessing the effectiveness of CAD. The radiologist can cover for CAD's deficiencies. CAD is not penalized for missing lesions as long as the radiologist finds them and CAD is not penalized for excessive FPs as long as the radiologist ignores them. The apporach works for expert radiologists, who know better than to mistake "silly" CAD marks for lesions (an actual characterization by a radiologist). However, it can fool the less experienced into initiating unnecessary recalls and even missing real lesions which were not cued by CAD and which regions the radiologist did not examine, because of over-reliance on CAD.
The second reader mode has set a low bar for CAD to be considered successful, and it should not surprise any one that CAD has "lived up to its low expectations". Massive clinical trials, involving about a million women, have shown that CAD is actually detrimental to patient outcome and there has been a call to end Medicare reimbursement for CAD. In the UK they do not use CAD. Instead every mammogram is interpreted independently by two radiologists resulting in improved performance. With expert level CAD the same could be accomplished in the US, with CAD being the second independent expert reader. Hence the name ExpertCAD™.