If you believe the hype, artificial intelligence is set to revolutionise dentistry, but there are barriers to overcome before the technology goes mainstream. By Angela Tufvesson
It might sound like the stuff of science fiction but artificial intelligence—or ‘AI’—is very real and proponents say it’s set to change the way dentists work in the not-too-distant future. Detecting cavities in X-rays, analysing dental images to identify problems and streamlining orthodontic visits are just some of the tasks emerging AI technologies promise to take off dentists’ hands.
But making the leap from start-up lab to clinical practice is a slow process. The dental industry is struggling to build the data banks AI needs—unlike what’s happening in medicine—and regulatory approvals processes can be maddeningly slow. So, what does the future hold for AI in dentistry: runaway success or frustrating stagnation?
When many of us think about how AI might impact the dental industry, we imagine robots taking the place of practitioners, rendering them obsolete. Thankfully, that’s not the case. This is ‘artificial general intelligence’, a level of computer intelligence that matches—not exceeds—human intelligence.
Instead, AI refers to computer systems that can ‘think’ and act like humans. They can perform tasks or imitate behaviour that would normally require human intelligence like speech recognition, visual perception and problem solving—but, crucially, these systems rely on humans to program or ‘train’ them first. The other key ingredient is data, and the enormous growth in the amount of high-quality data that’s generated and stored by the world’s computers is helping us realise the potential of AI.
What’s called ‘machine learning’ gives computers the ability to improve and ‘learn’ over time. It’s why Facebook gets better at identifying your friends the more you tag them. And it’s where the great power of AI lies.
AI in practice
When it comes to healthcare, there’s enormous potential to harness AI to deliver higher quality care with greater efficiency, particularly in early detection and diagnosis. In dentistry, some of the most interesting developments are in X-ray film recognition, image recognition and orthodontics.
LA-based start-up Pearl has developed a computer vision platform that it says reads X-rays and instantly and reliably identifies dozens of common pathologies. “Unlike other medical fields, dentists wear many hats, most notably serving as their own radiologists,” says Josh Tabak, vice-president of Product at Pearl.
“AI never gets tired, hungry or has personal issues affecting its ability to analyse a radiograph, reducing the risk of missing a critical diagnosis. This means less stress, more time, more money [for dentists] and happier, healthier patients.”
Closer to home, an image recognition tool called SmileMate is being used by Australian dentists to help patients. Users attach a special device to their smartphone, download an app and perform a one-minute screening of their teeth. The screening is analysed by AI and within minutes a report is generated, which outlines a detailed list of observations covering gum health, teeth health and teeth alignment.
“This tool gives you a way of keeping your patient engaged,” says CEO Nick Duncan. “For example, rather than calling a patient every six months to schedule a hygiene appointment, you can send them an email with information about what was observed last time, ask them to take a new set of photos, let them know there’s some new calculus build-up and that they should book in.”
In orthodontics, AI technology analyses huge banks of data to devise treatment sequences for sequential aligners, which helps dentists improve accuracy and efficiency. “Companies like Invisalign have these huge databases,” says Professor Heiko Spallek, head of school and dean at The University of Sydney School of Dentistry. “Invisalign has 50 million patients using it, and you can learn a lot from 50 million sequential scans—you see how the teeth are moving, you see what’s working and what’s not.”
Beyond the hype
But there’s a long way to go before these types of technologies become commonplace in Australian dental practices.
One of the biggest barriers to uptake is that the industry doesn’t hold vast banks of data, which are crucial for large-scale implementation of AI. “We are not that quick in translating AI into clinical practice because the underlying thing with AI is you must have a lot of data, and by definition you don’t have a lot of data in a single dental practice,” says Professor Spallek. “Compare that to hospitals, which have a lot more data.”
Another problem is inconsistent diagnostic codes. “In dentistry, we don’t write down diagnostic codes, and in order to have a treatment guided by AI you need to have some standard terminology for how you code diagnoses and treatments,” says Professor Spallek.
What’s more, gaining TGA and FDA approval for AI technologies isn’t easy—or cheap. It costs more than $70,000 per application for TGA approval in Australia and US$250,000 per application for FDA approval in the US, which is a lot of money for a start-up. Indeed, Pearl is still working to achieve FDA clearance.
Beyond these barriers, the industry must work through other practical considerations like who’s responsible for errors made by AI, how much AI should cost, and whether patients should be allowed to use AI directly without going through a dentist.
That said, experts agree it’s important for dentists to prepare for wider-scale implementation of AI. “Soon we will have more data in dentistry as well as implementation of diagnostic terminologies—so it’s getting there,” says Professor Spallek. “We’re seeing an increasing corporatisation of dentistry and these companies will have big data stores.”
He says the biggest impacts on clinical practice will be in risk reduction for medical complications and the implementation of AI in radiographs. “Expect recommendations for radiographs to come in the next five years.”
Nick Duncan expects AI to hit the mainstream within 10 to 20 years. “Dentists will have a lot of options in terms of how they implement it and the different suppliers—there won’t just be one provider in each space,” he says.
So how can practices prepare for AI? Go digital and collect lots of data. As Josh Tabak says, “Practitioners who are still using film to take X-rays and storing chart data on paper are going to find a higher barrier to entry.”