Improving the detection of cancer in the screening of mammograms

Examination of figures from screening at several different centres, and from the incidence of interval cancers in those programmes, suggests that a single radiologist reading by him/herself would miss 16 to 31% of cancers detectable in the mammograms. This study investigates how the accuracy of breast cancer screening could be improved.

Draw a comparison between screening for breast cancer and other inspection tasks that have been studied in the experimental psychological laboratory to suggest two low cost fixes that might improve the accuracy of breast cancer screening.

(1) Radiologists should not work for more than half an hour at a time; (2) two radiologists double reading should inspect the same series of mammograms in inverse order; (3) there should be a clerical check to see whether the efficiency of detection decreases with time on task; and (4) a more sophisticated engineering of the screening procedure might be accomplished by assembling a library of mammograms for which the correct diagnosis is known. If these library mammograms were mixed randomly with those to be inspected in some suitable ratio, it would then be possible to provide the radiologist with immediate feedback on the accuracy of (some of) his/her diagnoses. From experimental studies of inspection tasks this should improve diagnostic accuracy. These recommendations, of course, require evaluation in field trials.