AI in Healthcare: Are we ready for data to examine us?

You've heard it from us before on this podcast, but we'll say it again. AI is transforming our world.

Depending on which market research you look at, AI in healthcare is already somewhere from a 14-21 billion dollar industry in 2023, which is almost double what it was worth just two years ago. By 2028, it's set to be a 100 billion dollar global industry, growing some 40% year-on-year. That's astonishing, even in the already skyrocketing AI sphere. In this episode, we'll be looking at a wide spectrum of expertise to get a sense of where the field is right now, what the future looks like, and some of the cool technologies which might fill it. We’ll be looking at the ways in which AI is making healthcare more efficient, and overcoming roadblocks, as well as examining the ethics of letting algorithms influence human outcomes.

We’re joined in this episode by Mike Woodacre, Chief Technologist at Hewlett Packard Enterprise. He starts by spelling out the origins of this explosive growth during the COVID Pandemic, which ushered in a new world of collaboration and inter-disciplinary use of AI and High Performance Computing to look at new vaccine options, as well as examine scientific research looking for patterns. He urges caution, though, in relying on AI solutions which haven’t been adequately trained in the locales they are being used in and so may not account for regional factors such as more or less common versions of a disease.

That’s something Andy Cachaldora, General Manager for Northern Europe at GE Healthcare, agrees with. They’ve seen an incredible expansion of AI tools not just in diagnostic machines, but also in making sure that every second of a healthcare professional’s time is being used wisely. For him, AI is about taking out the grunt work and uncertainty from running clinics, giving better outcomes all round. Again, however, he urges caution in the way AIs are trained and implemented, with poor data collection and poor planning a route to disaster.

The idea of good, global data sources to train AI is something that has inspired Joachim Schultze, professor of systems medicine from the German Centre for Neurodegenerative Medicine. In collaboration with HPE, he’s been working on a blockchain-based system of machine learning tools to analyse Leukaemia scans, which keeps the data in-hospital to ensure data protection, but sends the insights of the scans to dozens of other institutes worldwide to train their own machine learning algorithms. That’s ensuring that everyone’s AI is collaboratively being trained on the widest, global dataset possible, with no risk to patient privacy.
But where’s the human in all this? Well, right at the centre of it all. After all, any AI requires training, and the training in most cases is still provided by human medical experts, for use by their peers down the line.

And a fascinating new piece of research suggests that the reason AI imaging works so well is that the expertise of a dozen doctors looking at cases together – in clinic or when training Ais, are better than one. A kind of swarm intelligence or swarm learning experience. Rutwik Shah worked on the research at the Center for Intelligent Imaging, which found that by training with swarms of doctors, not only could inexperienced groups of junior doctors analyse scans more reliably than the best AI, they were as effective as groups of doctors with decades of experience. It’s fascinating work, which could revolutionise the way AIs are trained and behave, as well as changing the way scans are analysed.

It's a fascinating world. Come with us on the journey.

Citations:
00:57: https://www.apa.org/monitor/2022/01/special-burnout-stress

01:55: https://www.marketsandmarkets.com/Market-Reports/artificial-intelligence-healthcare-market-54679303.html?gclid=CjwKCAjw5_GmBhBIEiwA5QSMxOR1CRaYz_g_dcLRAd1aJwxb3tbPBUYcQ0l9mrrtZsRv93yUzWkJTRoCruUQAvD_BwE
https://www.statista.com/statistics/1334826/ai-in-healthcare-market-size-worldwide/

06:19: https://www.bma.org.uk/advice-and-support/nhs-delivery-and-workforce/pressures/nhs-diagnostics-data-analysis#:~:text=There%20is%20not%20enough%20diagnostic%20staff%20in%20the%20NHS&text=According%20to%20the%20Royal%20College,11%2C370%20additional%20staff%20by%202025.

30:28: https://www.nature.com/articles/s41598-021-90292-6