So, I really talked about the three sides of AI in neurology. I talked about why AI might be the big transformation we’re hoping for to improve neurologic care, I talked about all the barriers facing that and then finally, I talked about how we might bridge the pros and the cons to try and arrive at a realistic vision of AI in neurology.
For me, I think AI reaches beyond neurology to medicine across specialties and I think there are various places where AI could play...
So, I really talked about the three sides of AI in neurology. I talked about why AI might be the big transformation we’re hoping for to improve neurologic care, I talked about all the barriers facing that and then finally, I talked about how we might bridge the pros and the cons to try and arrive at a realistic vision of AI in neurology.
For me, I think AI reaches beyond neurology to medicine across specialties and I think there are various places where AI could play. I think that includes diagnostics, investigations, and treatment selection decisions, but it also includes clinical operations, it includes population health management, and it also includes R&D, so drug discovery, for example. Within medicine and within that scope of all those clinical pathways I’ve talked about, I think there are really push and pull factors that make us want to implement AI in medicine, and neurology specifically. And so, the push factors are really technology based. So, we’ve had a revolution in our ability to compute. We’ve had CPU and GPU power and availability really changing the availability of cloud compute. We’ve also had new data sources. So we’ve had digitization of electronic health records. We’ve had neuroimaging on PACS systems since the 90s. Digital pathology is coming online. And we’ve also got new data sources. And I mentioned OCT in my talk, I also talked about Neurobooth at the Mass General Hospital, which is used on a research basis, as well as mobile health and wearables. And then finally we’ve got the icing on the top of the cake, if you like, which is AI, the ability to analyze that data.
The pull factors I talked about were really firstly cost based, that we’re coming to a cost crunch in medicine and can AI help with that? And then the non-cost factors, so patients rightly have increased expectations of us as doctors, and we want to personalize care more to them. So AI gives us potentially an opportunity to do that. In neurology in particular, I talked about how the brain is really complex, how data that we have from existing digitized modalities, plus new modalities coming online really helps us to capture that complexity. And then how we can use AI to better model that complexity, using high fidelity machine learning to better predict what’s going to happen to patients.
In my talk, I spoke a lot about data biases, and there’s many different types of data biases. There are data biases in terms of where country-wise we get our data from. Within countries, there’s data biases about which systems we get our data from, which brokers we get our data from, which geographies within countries we get our data from. And then there’s questions about looking at socioeconomically disadvantaged groups or different ethnic minorities, and whether we adequately and comprehensively capture data that’s relevant to them. So, there’s lots of sources of data biases, and there’s lots of thinking we need to do about how we combat them.
So, it’s something I didn’t get to mention in my talk, but I think data security and data privacy is a real problem. We only need to look at WannaCry, which happened a couple of years ago in the NHS. In the States, there have been lots of cybersecurity attacks. I think that we really need to think very carefully about security. We can’t afford for IT systems to be thought of as second-class citizens when it comes to health care. If we’re going to go with the data and AI revolution, we need to invest wholesale in making sure our data is private and secure. Because after all, if your supermarket gets hacked and gets your consumer data in terms of how you’re spending your money on food and other items, that’s one thing. You can change your password. But if your genome gets stolen, you can’t change your genome.
