The main problem, as I saw it, was that…

The main problem, as I saw it, was that radiologists do not have the chance to practice their readings over and over again, getting accurate feedback with each attempt. So this is what I suggested: You'd start by collecting a library of digitized mammograms taken from patients years ago along with enough information from those patients' records to know the ultimate outcome whether there actually was a cancerous lesion present and, if so, how the cancer progressed over time. In this way we would collect, in essence, a number of test questions in which the answer is known: Is cancer present or not?

— from Time

Also belongs to

Related