Of all the technological advances made this century, artificial intelligence (AI) is arguably the scariest. In just a few short years since AI became ‘mainstream’, we’ve seen several examples of its use for nefarious purposes.
One example came in April, with AI used to generate a fake vocal impersonation of science broadcaster Dr Karl Kruszelnciki. Dr Karl’s image and voice were used to promote a scam health supplement on Facebook. Incredibly, when he attempted to have the ads removed, he was told they did not breach advertising standards.
That’s one example of many, but AI can be – and has already been – used for virtuous purposes, too. One area in which this is already happening is health science. One of the newest examples is a blood test that draws on AI to predict the onset of Parkinson’s disease.
Incredibly, it may be able to make that prediction as early as seven years before any symptoms are displayed.
How can AI predict Parkinson’s?
The world’s ageing population has driven an explosion in cases of Parkinson’s, making prevention and/or cure a high medical priority. Of the two options, prevention is far preferable, something that relies on an early diagnosis.
Parkinson’s disease was first medically described as a neurological syndrome by James Parkinson in 1817, so science has had more than 200 years to learn about the condition. We now know that it’s caused by the build-up of a protein called alpha-synuclein. Alpha-synuclein damages or destroys nerve cells that produce dopamine in part of the brain called the substantia nigra.
Dopamine replacement therapy can help treat Parkinson’s various symptoms, such as tremors and muscle stiffness, but its success is limited. An early detection of alpha-synuclein build-up could potentially ‘head Parkinson’s off at the pass’.
With this in mind, scientists at UCL and the University of Göttingen set about harnessing AI to aid detection. They used a machine learning algorithm to spot a signature pattern of eight blood proteins in patients with Parkinson’s.
The algorithm was then used to successfully predict future Parkinson’s in other patients who provided blood samples. In one patient, AI correctly predicted the disorder more than seven years before symptoms arose.
As impressive as that sounds, detection at an even earlier stage remains a realistic proposition. “It is possible that it could go back even further,” said Dr Jenny Hällqvist, from the UCL Institute of Neurology. Dr Hällqvist is lead author of the study, which was published last week in Nature.
Is artificial intelligence the future of medicine?
Such was the success of the study, researchers are hopeful the test could be made available to Britain’s health service within two years. This will, of course, be dependent on successful trials with a wider population, but the outlook is promising.
One specialist has advised caution, however. Professor Ray Chaudhuri, medical director of the Parkinson Foundation International Centre of Excellence at King’s College London, said “major challenges” remain.
“Parkinson’s is not a single disease but a syndrome and can present in various different ways,” he said. “As such, management differs, and one size does not fit all. The prediction is unlikely to signpost these subgroups at this stage.”
Prof. Chaudhuri’s hesitancy is understandable, but the UCL study is undoubtedly a step in the right direction. It is also the perfect example of AI being used for good, rather than evil.
Do you know someone who has been diagnosed with Parkinson’s? What treatments have been made available to them? Let us know via the comments section below.
Also read: Parkinson’s drug shows promise
Disclaimer: This article contains general information about health issues and is not advice. For health advice, consult your medical practitioner.
