The UK desperately needs a Radiology Artificial Intelligence Incubator

To say that radiology in the NHS is drowning in work volume is an understatement. In May 2016 the Royal College of Radiologists highlighted the results of a national workforce survey, including the fact that 230,000 studies a month were stuck in a backlog of over 31 days and that spending to reduce these backlogs had increased by 57%. The story is the same in Scotland, where over £5.25 million per annum is spent on outsourcing. No-one seems to have the answer to this crisis, and with the current state of NHS underfunding it doesn’t look likely that one will emerge any time soon.

Over the pond, however, there is a huge push towards solving radiology’s problems — and the UK is being left behind.

Artificial Intelligence (AI) is being touted as the next big thing in analytical medical sciences such as radiology, with potential to not only drive down costs of reporting, but to ease the burden on overwhelmed radiologists while simultaneously adding quantitative data to standard reports.

Most of this drive has come out of the US, largely due to the capital infrastructure of a private payer system. In the past few years there has been massive stateside capital investment and institutional collaboration to further this vision. Some examples:

• In May 2016 Massachusetts General Hospital announced a partnership with NVDIA (the manufacturer of GPU processing units that are required to run large-scale AI) aiming to “improve detection, diagnosis, treatment, and disease management”.
• Then in May 2017, Mass Gen also announced further partnership with GE Healthcare on a 10 year plan to integrate ‘deep learning’ into clinical practice.
• IBM Watson is famously partnered with both the Memorial Sloane-Kettering Cancer Centre in new York, and the Mayo Clinic (who are also in cahoots with NVDIA).

Even more promising than these individual partnerships is the American College of Radiologists announcement of a new Data Science Institute. The DSI plans to work alongside government, industry and others to develop and implement AI. This comes as a welcome offshoot from the ACR clinical informatics push towards standardised DICOM, use appropriateness and workflow enhancement projects, and should help researchers and industry in validating AI in a standard regulatory and ethical framework.

It’s not just the US leading the way. Israel is proving to be an excellent incubating ground for radiology artificial intelligence companies. For instance, Zebra Medical is partnered with the University of Virginia, the University of Calgary, Carestream and Dell, and has now expanded to Australia. Tel-Aviv based AIDoc recently raised $7 million for it’s company, and MedyMatch raised $2million.

So why aren’t we seeing the same investment in the UK?

What we are lacking in the UK is the ability to attract these big name partnerships, and the funding and resources that come with them. The NHS is perhaps quite correctly seen as being too difficult to embed into, too complex and too slow. To bring artificial intelligence in radiology to fruition it requires huge databases of anonymised, labelled imaging data, with links to Electronic Health Records, on which researchers and tech start-ups or spin-outs can train and validate algorithms. However, the NHS does not have the funds to support this kind of work. (Let’s be honest, it doesn’t even have the funds to support it’s current workload!). Hence, this kind of big data warehouse simply does not exist within our shores. I find this shocking, especially since this idea is far from new — for example RadBank was designed and built back in 2008 at Stanford, and the RSNA has announced full support for a Quantitative Imaging Data Warehouse (QIDW) using federal funds.

The RCR has made some tentative steps towards considering artificial intelligence, by holding a forum in 2016, and word on the ground is that the Clinical Radiology Journal has commissioned a think-piece on radiology artificial intelligence, however progress overall has been at a snail’s pace. In the UK academic circuit there are dozens of medical imaging researchers building algorithms on small datasets, but they lack the resources to test them on millions of images, let alone get their product into the market. This bottleneck is choking British innovation.

What is needed is the alignment of big technology companies, the RCR and the NHS governing bodies to drive a fully collaborative vision in the field of radiology artificial intelligence. We should be capitalising on the NHS as a national system, by pooling imaging data and building a nationalised imaging warehouse and technology incubator (I’d like to call this BRAIN — British Radiology Artificial Intelligence Network). This would create a national institute for radiology in AI, capable of attracting industry partners, funding for researchers and equipment. We would then be creating fellowships for radiologists, building companies and spin-outs and partnering with PACS and hardware providers to implement the technologies as they emerge. Instead, we in the UK are left floundering, with scattered non-connected researchers trying to build products in silos, watching the rest of the world race away ahead of us.

Imagine an incubator environment with access to billions of NHS imaging records, where researchers can build, launch and grow technology start-ups and ventures. Imagine what can be achieved if, instead of training algorithms on just one local set of data, we could build algorithms on vast national data sets. Imagine if the funding and resources were in place to allow for experimentation, validation, growth and scaling. Imagine if regulatory approval processes in AI were baked-in to the incubation environment, so new products could launch straight to market.

Whether you believe in AI’s promise or not, the scale of investment and speed of progress across the globe surely indicates that many others do. If the UK is too slow to adopt a centralised model which actually drives innovation we will end up only being a customer; buying software from abroad which was built on the exact type of incubating environment we are failing to deliver.

If you are as excited as I am about the future of artificial intelligence in radiology, and want to discuss these ideas, please do get in touch. I’m on Twitter @drhughharvey

If you enjoyed this article, it would really help if you hit recommend and shared it.

About the author:

Dr Harvey is a board certified radiologist and clinical academic, trained in the NHS and Europe’s leading cancer research institute, the ICR, where he was twice awarded Science Writer of the Year. He has worked at Babylon Health, heading up the regulatory affairs team, gaining world-first CE marking for an AI-supported triage service, and is now a consultant radiologist, Royal College of Radiologists informatics committee member, and advisor to AI start-up companies, including Kheiron Medical.