Please provide a sample: A pathology lab approach to patient data, predicts expert

Deakin Software and Technology Innovation Laboratory is applying machine intelligence to healthcare

An AI expert has forecast the rise of ‘data labs’ in the healthcare sector, that will do for data what a blood lab does for blood samples.

The availability of sensors and wearables was leading to an increase in the amount of data that could be collected on patients, but doctors and GPs were not always able to make use of it, Professor Rajesh Vasa told the Digital Healthcare conference in Sydney yesterday.

“Most GPs are not trained to decipher a month’s worth of heart rate. They don’t know how,” the deputy director of Deakin Software and Technology Innovation Laboratory (DSTIL) said.

“They know how to understand 15 minute's worth but this is a month’s worth of data to make sense of.”

Just as a blood sample is sent to a lab for testing, a patient’s vital statistics – potentially heart rate or blood sugar levels picked up by a wearable device – could be sent or streamed to a specialist medical data lab for analysis, Vasa said.

“Currently there is no such profession,” he said. “And that profession needs to be invented, constructed, trained. And we somehow have to make it all work. That’s going to happen in the next decade.”

Despite the increase in patient data collection, the medical profession is yet to use the huge amounts of data effectively.

“There’s no sophistication. It’s just data collection for the sake of data collection. Only because it’s cheap to store it do we do it. There’s no real purpose for it yet,” Vasa said.

Warning system

Vasa shared some of the work DSILT was doing in the healthcare space, which makes up around a quarter of its current research.

One of the lab’s teams was working with the Royal Melbourne Hospital to develop a non-invasive method that used AI to predict an epileptic fit.

“The challenge is can we do it 30 seconds before it happens? If we had people wearing the head [electroencephalography EEG] helmet we can. What we’re now looking at is seeing if we can make a wearable do it,” Vasa said.

“There’s a long way to go. This is still an evolving field. But this will be available in the Fitbit you will get in ten years’ time as a standard feature. Wearables are not quite there yet, they’re not medical grade. They are designed for fitness freaks and work really well for that, but for the broad population they’re not that useful.”

Although the warning would come only a short time before a seizure, the benefits were significant, Vasa said.

“It’s not enough to do serious stuff, but maybe they can sit down and that would mean less brain injury and head trauma. We can reduce the chance of death quite a lot just by telling people to sit down or, if you’re driving, just pull over."

Vasa admitted that although the technique had an 80 per cent likelihood of correctly predicting a seizure, for some people it didn’t work at all and for some types of seizures it didn’t work either.

In December, researchers at DSTIL won an innovation competition run by the The Epilepsy Foundation SUDEP (sudden unexpected death in epilepsy) Institute, with a related piece of research that identified physiological patterns present in patients during and in-between seizures.

In a separate effort late last year, the University of Melbourne posted patient EEG data on data science community platform Kaggle in order to crack the “holy grail of epilepsy research — an algorithm that can predict seizures”.

Doctor knows best

Although AI will be a useful tool to doctors, the technology will remain just that for some time, Vasa said.

“'Will it replace me?' is a question I always get asked by healthcare workers,” he said. “If you’re a healthcare professional, you’ll be way past, you’ll be retired, before this gets even close to thinking of it. So don’t worry about it’s not in your lifetime it will probably not be in the lifetime of your great grandchildren.”

Join the Computerworld newsletter!

Error: Please check your email address.

Tags healthAIepilepsyDeakin UniversityHealthcareartificial intelligencemachine learningDeakin Software and Technology Innovation Laboratory (DSTIL)

More about Royal Melbourne HospitalTechnologyUniversity of Melbourne

Show Comments