R&D is critical in the world of healthcare, but it is also tricky; progress is slow, and further progress is hampered by the way different researchers and drug developers work in silos, data is kept secret, locked away from the rest of the world. Can AI in healthcare come to the rescue?
Vas Narasimhan, CEO at Novartis, recently warned of a problem finding new data. In an interview with Bloomberg, he said that this lack of data has in part caused his initial enthusiasm for AI to turn more cautious.
This may yet prove to be the single biggest hurdle in applying AI in healthcare to help find cures to new diseases, extend life, and improve the quality of life.
Julien de Salaberry, CEO of Galen Growth Asia, a Singapore based organisation that is creating a $70 billion health tech ecosystem across Asia put it this way: “AI as a tool to aid research is still largely hype.”
He explains: “Its use is hampered largely because of the way data is held in silos and the interoperability of data is not there — everybody is hoarding data as if it was gold, and so the academic centres don’t speak to each other, the researchers don’t speak to each other, either.
“Everyone is building data, it’s growing faster than any other commodity on the planet, but no one is really sharing that data.”
Blockchain and privacy: Can a form of distributed ledger solve the problem of privacy?
He reckons that thanks to GDPR, the sharing of the data is even harder.
More is the pity, because AI in health care really could be a big deal.
As Narasimhan put it when he was speaking at the World Medical Innovation Forum earlier this year: “As we try to solve the puzzle of new drugs, we are really trying to be molecular locksmiths, unpicking biology that evolved over two billion years. For most of human history we have been unable to solve that puzzle.”
AI in healthcare, the opportunity
Even now, 100 years or so from the inception of the pharmaceutical industry, Narasimhan says: “We only have around 1,500 molecular entities that have ever been licensed, we only have about 400 to 500 tractable targets. Most of human biology is still not understood.
AI and machine learning needed to improve the healthcare industry
“So, what I wonder is… can artificial intelligence, can a series of very bright start-ups, change the possibilities of how we understand human biology, and unlock what are probably another million targets, that are out there, but we don’t understand.”
That’s quite the challenge we are setting AI in health care, but the prize is great indeed. So far, we have found 400 to 500 tractable targets, AI may unlock another million. It is like we have just heard the starting gun in a marathon, the ground we have covered, compared to what we could cover, is enormous. The problem of lack of sharing of data, or of it sitting in silos, has to be fixed.
Part of the opportunity lies with finding research. De Salaberry invites us to “imagine being a scientist with access to all the data you wanted in the world on one specific disease area, you’d probably find that an awful lot of what you wanted to do was already done. AI, combined with data, could be able to fine tune your hypothesis.
This is why China, where data is more freely available, could have such an opportunity. De Salaberry says: “One of the reasons why China is probably going to win the AI, let’s say, arms race, is because their approach to data is very different to the European and US approach.”
Being a CTO for a healthcare technology company
To illustrate the point, he cites the examples of various genome research projects. “In order to be effective in genomics, you need a database of genomic data to make comparisons. So, the US has declared it’s building a dataset with one million individual data subjects, with information on each individuals’ genome, in order to have a database that’s sizeable enough to draw meaningful inferences.“
So that’s a million in the US. By contrast, Singapore, with a population of less than 2% of the size of the US, “is building and analysing a database of a similar size. China is building “a database with 100 million data subjects,” says de Salaberry.
On the other hand, there are other ways AI and healthcare can dovetail.
Explaining, de Salaberry, says that one of the benefits of AI in healthcare could be to help a nurse or a doctor “accelerate their arrival at a diagnosis. Not only could it help compute the data points you have on a patient faster, it could also be able to compare it to a broader database as well and then provide the analysis for the doctor or nurse to apply. The interesting thing about that data is that some of the early wins relate to proof that a healthcare professional, armed with data provided from a wearable device, actually enhances the accuracy of diagnosis. So, by itself, the AI has a certain precision level; the physician will have their own precision level, but this could be quite variable depending on the individual. Put the two together and now you’ve actually enhanced both to a high level of precision.”
Or as Narasimhan said: “Our odds at Novartis of finding bad decisions, then making the right decisions, go up when we are powered by these machine capabilities and artificial intelligences.”
AI and machine learning needed to improve the healthcare industry
