"The researchers started with a Google-developed algorithm primed to differentiate cats from dogs."
-Yep, and you can do this at home:
http://adilmoujahid.com/posts/2016/06/introduction-deep-learning-python-caffe/
Great articles overall, worth some discussion. I believe medicine will be profoundly affected by machine learning (for the better). I'm just not sure how the course of adoption will play out. There are basically two paths:
1. Departments adopt machine learning (i.e. your radiology or dermatology department purchases the technology). I think this will be the more likely pathway and is
already happening in pathology. Pathology is incredibly well positioned for this - they already use computers to incubate cell lines and perform lab measurements. More importantly, the leadership structure of the modern pathology lab is amenable to machine learning. Pathologist oversee pathology technologists who perform the initial read. If you've ever been in an IR or MSK biopsy, you've probably run into the path tech who exams the slices and determines if more tissue is needed and makes an initial call. The initial reads can be done by machine learning.
2. Referring providers adopt the technology - This is the less likely path, but the most dangerous for radiology/dermatology departments. Take melanoma as an example. If a primary care provider can use a machine learning tool to identify melanoma, he or she can directly refer to a general surgeon for surgical removal. This avoids the standard referral to a dermatologist. A similar scenario is even more likely in radiology as providers routinely read their own images and use the radiologist's read as confirmatory.
How it plays out is ultimately up to the leadership of each institution. FDA approval usually signals an avalanche of change. Overall, the Wired article hits the nail on the head:
"The key to avoiding being replaced by computers, Topol says, is for doctors to allow themselves to be displaced instead. 'Most doctors in these fields are overtrained to do things like screen images for lung and breast cancers,' he says. 'Those tasks are ideal for delegation to artificial intelligence.' When a computer can do the job of a single radiologist, the job of the radiologist expands—perhaps to monitoring multiple AI systems and using the results to make more comprehensive treatment plans. Less time drawing on X-rays, more time talking patients through options."
