Computer aided detection
- PM Taylor1
© Current Science Ltd 2000
Published: 1 October 2000
Computer scientists first started thinking how computers could be programmed to interpret mammograms in 1967. The field (known as computer-aided diagnosis or CAD) became increasingly active in the 1990s with the promise of digital mammography. The research continues, but the best of the algorithms developed over the last decade have now been incorporated in commercial systems. These algorithms are capable of high degrees of sensitivity only at relatively low levels of specificity. The systems that depend on them are therefore best used to prompt a radiologist (or other trained film-reader) to the possible presence of an abnormality.
Investigating the potential of prompting systems means addressing questions about how technology affects the way people work: how does prompting affect screening performance? How much information can film readers make use of? Does it make new models of screening cost-effective? The challenge is to answer these questions within a timescale appropriate to the pace of technological change. We expect, over the next 18 months, to conduct an NHS-funded trial of the R2 Imagechecker. Economic models of the screening process will then be developed to compare the cost-effectiveness of prompting in a variety of screening protocols, including film reading by non-radiologists.