Artificial Intelligence and Healthcare

We trust Siri to send a text and perhaps order a pizza – but do we trust her to prescribe our medications?  Let’s take a quick peek at how AI is working in our hospitals and at the doctor’s surgery.

Your doctor may have already subjected you to the Australian designed smart stethoscope, but AI is having impacts across a broad range of healthcare areas.   An ageing population combined with the availability of more expensive treatments, means getting value for our healthcare spend is a priority for governments around the world.

Photo by Franck V. on Unsplash

X-rays and Scans

At the heart of Artificial Intelligence is the concept of machine learning.  And with machines, when we say “learn”, we mean “gets better results over time when given more examples”.  Detecting disease in lung CT scans can be laborious. Learning from thousands of examples of both diseased and healthy scans, AI leverages modern day algorithms and processing power and combines them with statistical models.  These can make light work of detecting which scans from new patients are likely to be diseased. This can free up radiologists’ time from routine tasks, allowing them to concentrate on the more difficult cases needing an expert hand.

In the US, the FDA has approved only one AI product for any application so far.  The IDX system is used for identification of diabetic retinopathy – damaged blood vessels in the eye associated with diabetes.  Once the disease is detected, the patient is referred to a specialist for treatment.  The manufacturer claims that the IDX system has reached the same accuracy as a specialist at detecting the disease.  Not everyone accepts this statement, with concerns being voiced about the validity of small sample sizes and questions over results for patients with multiple eye diseases.

By Kiatdd via Wiki Commons

On-line service

London has trialled an AI assisted chatbot to perform triage for non-emergency users of their ”GP at Hand” smartphone app, which aimed to reduce hospital and doctor visits by providing clinical advice directly.  The vendor claims it can achieve diagnostic accuracy as good as doctors.  Proving popular and convenient, thousands of people have consulted the app.  Rwanda also uses the app to provide cheap healthcare to its entire population.  

But not everyone is happy to trust it.  Wired magazine rated it third of three health apps, finding issues within minutes.  London has commissioned an independent evaluation of the app and its usage.

View from the US

The US government is optimistic about AI applications in both the above areas of imaging and online services and also more broadly.  Advances in cardiac imaging have the potential to avoid the cost and risk involved with invasive diagnostic techniques.  With the right processes in place to validate AI solutions and to provide the right data to adequately train the systems, a large cost saving and benefits to the community are possible.  

Australia and the future

The CSIRO recently produced The Future of Health report, a look at the health of Australia as a whole.  We will need to leverage advances in AI for diagnosis and decision support, as well as find the right balance of regulation to protect patients while encouraging investment if we want to get the most out of this technology.  If we don’t pursue it, we will be buying AI healthcare systems from countries that do.

7 Responses to “Artificial Intelligence and Healthcare”

  1. Your post raises quite an important point, even despite apparently improving accuracy SHOULD we be leaving this sort of thing to AI? Maybe it would be better placed as a support tool?

    It’s almost like half of people are TOO excited about AI and won’t give it the time to mature as a technology, and the other half are in complete fear of it! A lot to think about!

  2. Cynthia says:

    Thanks for your comment. There is so much promise in this technology – I hope we can do it right the first time and smooth the way for the mainstream adoption of AI.

  3. Seana Glover says:

    Great article, Cynthia! Very informative and intriguing. I can imagine it may be a while before AI becomes an integral part of the Australian healthcare system, but it is definitely an interesting and necessary hope for the future!

  4. Cynthia says:

    Thanks for your feedback – it’s a treat to hear from somebody at the front line. Hopefully some early AI systems will gain public acceptance. It took 17 years for robotic surgery to become mainstream, hopefully we can beat that with AIs.

  5. Zoe Canestra says:

    The triage example would be such a huge help I think! I am a medical receptionist, and the amount of patients we have that should have gone directly to hospital or, on the other end of the spectrum, should have just stayed home and rested, is astounding. Online systems free up our time and the doctors time for more important things, like online booking systems for appointments mean that only the people with questions for doctors or complex needs call us! AI could be huge for the medical industry. Love the article!

  6. Cynthia says:

    Thanks Alexander. Although Siri and Google Assistant are getting better every year, they still have a way to go.

  7. Alexander says:

    Ahhh, I loved your way of explaining machine learning as not “learning” but getting better with more examples. Although this might be argued as learning, it’s a clean way of detailing the fact a computer can’t learn the same way us humans do.