May 31, 2019
Can AI succeed in healthcare?
The possibilities are endless but we need to get over our trust issues, says Dutch physician and researcher Martijn van der Meulen.
We may not want to think about AI (Artificial Intelligence) penetrating every area of our lives, but truth is, it already is. “Do you use Google Maps, Waze, or predictive text on your Smartphone?” asked Martijn van der Meulen in his presentation to HealthPRO senior leadership and board members. “It may not be the way you imagined it, but you are definitely already using AI.”
Martijn van der Meulen presenting to HealthPRO's Board of DIrectors and senior leadership team.
This is what van der Meulen, MD and researcher at the Radboud University Medical Center Nijmegen in the Netherlands, says is the essential dichotomy in AI – fear versus dependency. “There are so many elements about AI that scare us – privacy, inequality, losing our jobs,” he explained. “But, at the same time, we’re using it every day and couldn’t imagine living without it.”
In his talk, entitled Artificial Intelligence in Healthcare, van der Meulen tackled the elephant in the room inside many healthcare organizations – if so many companies are investing in AI, why hasn’t healthcare stepped up? Theories abound, many of them rooted in his follow-up question, “How will AI succeed in a field that is so human-centred?”
From the floppy disk to big data
When it first arrived in the 50s, machine learning held wonderful promise of transforming many seemingly old school industries. As progressive as it was, machine learning stalled because we didn’t have the data or computing power. It re-emerged as a force in 2010, says van der Meulen with the development of big data and the internet of things (our technological devices). “We forget how these things evolve exponentially,” he says. “In just a few years we went from the floppy disk with limited storage to big data and cloud computing with unlimited data storage.”
Van der Meulen said that AI is making things happen in every industry at a dizzying clip, and not surprisingly, it’s having an impact on the global workforce. Fourteen per cent of the current workforce’s tasks are highly automatable and 6 out of 10 current occupations have more than 30 per cent of tasks that are technically automatable. We can expect most jobs to change due to AI, but Van der Meulen reassured us that its implementation will result in an increase of jobs overall. Afterall, he said, “who could have predicted that there would be an entire job category devoted to managing social media platforms?”
Humans and AI have many parallel capabilities. “We rely on algorithms, which are nothing more than a recipe that helps us get through every day,” he says. What connects us to technology is the fact that we are both always making task-based decisions and anticipating what we’re going to do next. Where the parallels end is in the area of moral reasoning. “Despite its myriad capabilities, as a tool AI is still not infused with morals and values,” says van der Meulen. “Can you infuse those values? Sure, but they will be tainted by developer bias.”
Which could explain why healthcare is so late to the AI party. “Our business is rooted in empathy, communication, critical thinking,” he said. “We still don’t think machines have those kinds of ‘human skills’.” But to disprove that point and highlight the incredible possibilities presented by AI in healthcare, the young physician presented the example of his own country.
Van der Meulen says many hospitals and healthcare organizations in the Netherlands & the UK are using AI for the complex problem-solving and digitization of processes. These include infrastructure development, smarter personnel scheduling and predicting 911 calls by looking at regional hotspots. “AI is being used to help us navigate hotspots where higher numbers of cardiac arrests are taking place, which enhances our predictive capabilities,” he said. “The day might not be too far away where the ambulance will arrive before the heart attack even happens.”
AI is also making advances in procurement in the Netherlands – specifically with spend classification, AI-supported sourcing, supplier risk identification and big data benchmarking.
Robots: coming to an operating theatre near you
Hopefully being human will always ensure us a place in healthcare, but on the important scorecard which is inevitably the numbers, AI will translate into real dollar savings. Van der Meulen cited ten AI applications that could change healthcare, and at the top of the list is robot-assisted surgery that could save $40 billion by 2026. “There is so much potential,” he says. “In the operating room, a robot can do in 20 minutes what it takes a neurosurgeon two hours to do.”
Although the savings on the administrative workflow side looking encouraging ($18B), van der Meulen doesn’t deny there are still a number of important obstacles to overcome. “These systems, known as deep learning, need large amounts of data to be feasible, they don’t all talk to each other, and there are inherent governance and security questions.” As we’ve witnessed with the transition to electronic health records – bad data in, bad data out.
Deep learning is still at the top of the hype cycle
It might come down to AI and humans learning to understand each other better. “We’ve seen in different industries that the real potential of AI is with humans and working with people,” he says. Learning how to talk about it is a good first step. Van der Meulen presented a graph showing deep learning at the top of the hype cycle – in other words, it’s at the peak of inflated expectations. Not yet in the trough of disillusionment but not on the slope of enlightenment either.
“There are so many possibilities,” said van der Meulen, both on the logistical side (locations, planning, and workflow optimization) and non-logistical side (personalized medicine, prevention and risk assessment). But in the conservative and risk-averse culture of healthcare, it’s much harder to convince people that these new AI systems are safe and worthy of our trust.
If we can overcome our trust issues, AI could lead to amazing breakthroughs says van der Meulen. And we are well-positioned in healthcare to explore them. “There is resistance, so we need pioneers and champions to demonstrate the benefits of AI. Some jobs will not exist but more jobs will be created. This is a very positive development, but first we need ambassadors who can prove that doing something different is worth it.”