- Algorithm is better at identifying medical conditions than human doctors.
- Artificial Intelligence was fed over 100,000 images to train it to diagnose diseases.
- Technology could give better diagnostics with fewer errors.
Google has always been at the forefront of developing new technologies and has made quite a fortune out of it. Just this month, the company announced that its artificial intelligence is now better at diagnosing medical conditions than human doctors.
New Google algorithm can save thousands of lives.
While many are calling it another step towards a future robot uprising, the technology is a welcome addition to medical science. Thousands of people die each year due to incorrect diagnosis, something that can be prevented by using a thorough AI.
Google made the announcement in the Journal of the American Medical Association on Thursday. The paper outlined the company’s work with deep learning algorithms and how they can be utilized for medical purposes.
The AI that the company used was fed over 128,000 retina photographs to train it in diagnosing diabetic retinopathy. The resulting algorithm was then tasked to identify the condition in a set of 10,000 new images.
The results were astonishing, to say the least. The artificial intelligence was better at diagnosing the disease than real doctors.
Algorithm On-Par With Ophthalmologists
In a blog post in the company’s research blog, Google’s Product Manager Lily Peng MD PhD stated, “The results show that our algorithm’s performance is on-par with that of ophthalmologists.”
One of the most common ways in detecting the condition is to have a specialist examine pictures of the back of the eye. The specialist is then to determine if the patient has the condition and if so, rate its severity.
This method is highly dependent on the specialist which, despite being an expert, is still prone to human error. The introduction of an AI that can do such a tedious task could give way to a better diagnosis with a lesser margin for error.
The algorithm made by Google is still crude by technology’s standards and can only be used for diseases which rely on visual diagnosis. However, this is only the first step in the new age of medical technology. And as efficient as it might seem, it will still be decades before doctors can be fully replaced by artificial intelligence.
Spoke about this at I/O: today we're publishing our results of using machine learning to detect diabetic retinopathy https://t.co/k7oWRO04cK
— sundarpichai (@sundarpichai) November 29, 2016
— ❄John B. Roberts❄ (@auroberjo) November 29, 2016
"Diabetic eye disease is caused by… https://t.co/zX3N1lBR4B
— SEVA Skeptics (@SEVASkeptics) November 30, 2016
Also Read: Google AI System Now Considered A Driver