Exploring computerized mammographic reporting with feedback

摘要

Computer-aided diagnosis is generally dependent on feature detection and identification on digitized images. In this project, we explored the use of a computerized system in which features identified by radiologists for a previous set of mammography cases, in addition to other information on those cases, formed a database of information that provided readers of new cases with statistical probabilities (feedback) for malignancy in similarly rated cases. Although in this preliminary study the feedback system did not improve levels of diagnostic accuracy (which were already high), the results suggest that in certain settings and with certain mammography readers the method may be extremely promising.