While artificial intelligence (AI) is being applied almost everywhere, by far the most exciting field for my money is medicine. AI’s inhuman ability to absorb millions of data points and spot patterns that even the most trained doctor’s eye can miss is enormously exciting, and now there’s a very promising possible development in early diagnosis of the cruellest of diseases: dementia.
The BBC reports that an AI system capable of spotting the warning signs of dementia from a single brain scan has passed pre-clinical tests and is ready to be tested in the real world. And because the system compares scans to thousands of dementia patients on record, it may even be able to estimate the speed at which a patient’s condition will deteriorate, helping doctors decide what steps to take in order to slow the progress of the disease.
“If we intervene early, the treatments can kick in early and slow down the progression of the disease and at the same time avoid more damage,” Cambridge University professor Zoe Kourtzi told the BBC. “And it’s likely that symptoms occur much later in life or may never occur.”
In its pre-clinical trials, the system was able to diagnose dementia years before actual symptoms emerged, even when no obvious damage was visible to doctors via a brain scan.
About 500 patients are expected to be analysed by the system in its first year at the Addenbrooke hospital in Cambridge, as well as other memory clinics around the UK. The AI will be used alongside traditional methods of diagnosing the disease for comparison.
As exciting a development as this is, it’s important not to get too carried away as it’s still in its very early stages, and could yet prove to be a false dawn. As Professor Clive Ballard, an expert in dementia from the University of Exeter told The Guardian, “we need to be really careful not to create false expectations.”
This isn’t just your usual expectation management, either: dementia just manifests itself differently in patient brains. “AI has been shown to improve the diagnostic potential of brain scans compared to clinical reading of the scans, but there is so much heterogeneity between individuals that it is completely infeasible for a single scan, biomarker or clinical test to be that certain in a single assessment,” Ballard cautioned.
It is sometimes the case that artificial intelligence is able to spot trends between images that experts don’t think to look for, either because they’re considered irrelevant or simply invisible to human eyes. Hopefully that will prove to be the case with this trial, and sufferers can get more advanced warning, a better quality of life or, ideally, both.