AI software that helps doctors diagnose like specialists is approved by FDA
In a wonderful technological advance, artificial intelligence has been trialled and tested in helping doctors diagnose tricky patient cases. Whilst we are a number of decades away from total automation, this is the first big step that could lead to a more accurate reading of patient symptoms, fewer human errors and more comprehensive treatment plans in the future.
The advance has come in the optometry field with a program affectionately known as IDx-DR. According to founder Michael Abràmoff, it makes the clinical diagnosis on its own, working autonomously, rather than with someone constantly monitoring its output.
“It makes the clinical decision on its own.” he said.
The way it works is, a human, usually, a doctor or a nurse, in this case, uploads a high-resolution image of the patient’s eye. After determining it is able to get a read on this image, the software is able to recognise patterns that determine whether the patient has diabetic retinopathy or not.
The patterns have been recognised and developed through the software with the aid of thousands of images. But just how accurate is it?
At this point, the program is far more accurate than human doctors. In fact, 87% of the time it was able to correctly identify diabetic retinopathy in case studies. Also, 90% of the time it was able to correctly identify that patients did not have the condition. It has now gained FDA approval in the States – the first such advancement to achieve such.
IDx-DR is not the only medical AI program out there, either. Google is heavily into the research field through its DeepMind projects. Questions are being posed about the removal of a human element in medicine, for example, where do liabilities fall in the event of malpractice or false reads?
It seems the world is marching faster and faster towards total autonomy. One exciting offshoot, aside from higher diagnosis accuracy, is potential access to cheaper medical services down the track. Could this translate into cheaper insurance, and Online doctors?
Would you be happy with a robot telling you whether you had a life-threatening illness?
H/T: The Verge