How can AI help 3D-print human body parts?

15 Oct 2018

Axial3D team. Image: TechWatch

TechWatch editor Emily McDaid speaks to the team behind Axial3D about the role of AI algorithms in transforming 2D images into 3D prints.

Well-known Belfast start-up Axial3D produces 3D prints of body parts. This isn’t to satisfy the narcissistic social media types – it has important surgical implications. This previous TechWatch article describes the company’s process.

Now, Axial3D is developing new AI techniques to make instantaneous the transition from 2D images to 3D prints. How is it doing that?

CTO Niall Haslam said: “To produce our models, we take 2D images of slices of the anatomy to build up a 3D model. This process requires annotating the images to denote bone versus muscle versus organ. At the moment, this task is completed by medical visualisation engineers – we have a team of five of them – and it can take up to four hours per print. We’re now developing a machine-learning technique to automate this process and make results as fast as possible.”

It wasn’t clear to me why instant results were important, so CEO Dan Crawford explained it. “When you look at the scale of surgery, there are 3m operations per year. If we want 3D printed organs to be the norm, to unlock that market, we must automate this process,” he said. “Luckily, there are generalisable features in any person’s anatomy – in other words, commonalities with how their heart or spine will look. We can use previous scan files to train a classifier to label images.”

He goes on: “There’s no way 3D printing could become a routine medical procedure without automation.”

two young men sitting at table, one with red hair, a beard and glasses, beside a short-haired man wearing blue shirt.

From left: Niall Haslam and Dan Crawford. Image: TechWatch

I get it. This business is unique and hugely promising but, rather than being a ‘nice-to-have’ bit of extra medical attention that only wealthy patients could afford, to bring it to the mass market in all hospitals means making it instant.

To imagine a hospital of the future, a 3D printer could sit in the same area of the hospital as your MRI or CAT scans. You go in for a scan and come out with a 3D model of your body part.

As Axial3D grows – and it has done significantly in the past two years since I first met the company – I wonder whether it will expand into the medical implants business. At the moment, most of the 3D prints are used before surgery, so the surgeon can perfect the technique and know exactly the layout of blood vessels, bone and so on. Crawford said: “Implants could be a new focus for us but we need to perfect our AI algorithms.”

That’s why Crawford hired Haslam. His experience was eerily perfect for this job, since he’s worked in both biology and software engineering.

Crawford explained more about whether they could 3D-print implantables: “Everything is getting more cost-effective and there’s innovation in clever materials. Right now, we work with FDA-approved polymers but we’d look to move into metal implants – titanium alloys – over the next few years.”

Imagine if a person had a head injury and a portion of their scalp needed to be replaced. The injured area could be scanned, and the new piece of scalp 3D-printed so that it was an exact fit.

group of young workers sitting around a table looking at a presentation one is making using a laptop.

Axial3D team. Image: TechWatch

Axial3D has won a prestigious place with the NWCAM project for its AI development, and it has got four academics based at NIACE who will be working with what Crawford calls “clever materials”. That project will extend for a minimum of three years.

Crawford is quick to explain: “We’re not polymer chemists ourselves.” He said: “There are two stages to the process: the virtual stage, where a medical image becomes a 3D file –that’s our portion of the business; and then the physical process of turning it into a 3D object – this is what our partners do.”

This proprietary AI algorithm will be critical to Axial3D’s success. Has the company filed a patent? Haslam tells me: “We filed an initial patent in May to describe a lot of the AI specific parts. We also have an initial submission to describe the 3D native operations; how we manipulate objects in 3D.”

What are its biggest markets? “Orthopaedics, cardiology, oncology and neurosurgery,” Crawford said and, in terms of geography, the US.

Will it open a US office? “We’re doing work in San Francisco and Boston at the moment, and a Boston location would make sense due to the shared working hours with our HQ in Northern Ireland,” Crawford said.

Are they actively selling into the US market? “Hospitals are assessing the beta with positive results,” said Crawford.

What about funding? Crawford said: “We’re going through a significant Series A round right now.” That new funding will enable Axial3D to extend its reach far beyond the first installations it had this year, among them Royal Victoria Hospital, Barts at Royal London and University College Dublin.

Crawford sums it up: “The trials at Royal Victoria didn’t just improve the quality of care and saved a life – they also reduced hundreds of hours of costly paediatric care.”

By Emily McDaid, editor, TechWatch

A version of this article originally appeared on TechWatch

TechWatch by Catalyst covered tech developments in Northern Ireland