To be fair though, those numbers are the results of a single evaluation of the algorithm against a single group of human experts:
> In the setting of a challenge competition, some deep learning algorithms achieved better diagnostic performance than a panel of 11 pathologists participating in a simulation exercise designed to mimic routine pathology workflow; algorithm performance was comparable with an expert pathologist interpreting whole-slide images without time constraints. Whether this approach has clinical utility will require evaluation in a clinical setting.
Certainly an impressive feat of image recognition, but far from revolutionising cancer diagnosis. It's also not clear to me that this would actually diagnose cancer more accurately if you factor in the ability of experts to consider other things than just the scans.
I guess so though it seems promising. Not quite sure how it'd work in real life. I've got two friends who went to the doctor with a headache and sore throat and were told it was nothing to worry about and then a year or two later found it was cancer - one died rapidly, one presently having a bunch of surgery. It would be good if there were a better way of screening that sort of stuff.
> In the setting of a challenge competition, some deep learning algorithms achieved better diagnostic performance than a panel of 11 pathologists participating in a simulation exercise designed to mimic routine pathology workflow; algorithm performance was comparable with an expert pathologist interpreting whole-slide images without time constraints. Whether this approach has clinical utility will require evaluation in a clinical setting.
Certainly an impressive feat of image recognition, but far from revolutionising cancer diagnosis. It's also not clear to me that this would actually diagnose cancer more accurately if you factor in the ability of experts to consider other things than just the scans.