I find it most interesting to use tools like this to augment - not replace diagnosticians and specialists. Of course the key here is that the training set is c

I find it most interesting to use tools like this to augment - not replace diagnosticians and specialists. Of course the key here is that the training set is crucial to building a high quality model - which of course needs a set of specialists to give their consensus on the diagnosis of the patient based on the images. Presuming those folks can agree - the technology becomes a force multiplier for good. If they disagree or label things problematically - they become a force multiplier for bad.