The use of AI in diagnosis
In his book 2084, Lennox reports that at the University of Toronto, engineers have harnessed the powers of machine learning in generating X-rays which augment training sets.
Not only are these anywhere from at least 20% to as much as 40% more accurate in diagnostication than physicians, but because they are also computer generated – and not from actual patients – there are no privacy concerns in sharing these with researchers outside the hospital.
Dr. Shahrokh Valaee was quoted in the book as saying this,
“It’s exciting because we’ve been able to overcome a hurdle in applying artificial intelligence to medicine by showing that these augmented datasets help to improve classification accuracy…Deep learning only works if the volume of training data is large enough and this is one way to ensure we have neural networks that can classify images with high precision.”
Apparently, this is being used to identify diabetic retinopathy, among other difficult conditions to recognize.
Surgical Medical Imaging
In Hong Kong, they have designed a robotic system within an MRI scanner to perform neurosurgery. This is one treatment they will be using for Parkinson’s and major depression, among other afflictions.
How development of medical AI will save fortunes
The National Health Service (NHS) in the UK projects that if they reach their targets in AI development they will save as much as £1 billion by rendering some 30 million outpatient appointments unnecessary, allowing them to invest more into front-line care.
In the not too distant future, I will be discussing a book I’ve been invited to review entitled We Can Do Better by David Goldbloom. It was just published a few weeks ago on May 4th (I wonder if ‘May the 4th be with you!’ had an influence on the drop date).
In future blogs covering that work, I will be touching on how – despite certain alarming trends – health apps are expected to be relieving a lot of emergency services and spending, too.
At this point, it’s very difficult to establish quality control and proper regulation for these apps. But there are some very promising, and already extremely helpful programs where we have seen encouraging results in treating mental illness.
The popular downloads called Woebot and Wysa are perhaps the greatest examples that spring to mind. But there are others, like A4i, which administer a form of game therapy called cognitive adaptation training. This aids ones with various mental illnesses and learning disorders from managing daily organizational tasks (like scheduling their day or filing taxes) to improved social skills (like holding a conversation)…even dating.
Wearable AI for monitoring and early detection
We’re already seeing how wearables, like smart watches, are now primarily being used for health monitoring. According to John Lennox in 2084, one team at MIT, together with the company Empatica, have created the first machine learning system which uses a smart watch to recognize seizures. In fact, it’s already FDA approved and on the market in the US and the EU.
These are all wonderful advances and uses for Artificial Intelligence. There is much AI alarmism that is justified – and I’ve been detailing them quite extensively here on medium.com – but these are, on the whole, very positive developments. I haven’t even mentioned the half of it, either.
Tell me what you think of all these strides in the remarks below. Is AI mostly positive, such as what seems to be the case in medical uses of it? Or, based on some of the other concerns I’ve been journaling about of late, do you think it is Pandora’s box?