The AI Breast Cancer Plan - putting the brakes on innovation?

This week’s announcement of what was described as a “world leading” trial to test AI tools for examining breast cancer scans seemed like really good news. At last it feels as though the NHS is moving into the fast lane when it came to adopting artificial intelligence. The press release from the Department of Health said the trial involving 700,000 women and five different AI tools was “the latest example of how British scientists are transforming cancer care, building on the promising potential of cutting-edge innovations.”

Then as I began to ask around a bit, the picture became a lot less clear. According to one view, far from moving the NHS into the AI fast lane, the trial would be an interminable affair involving technology that would rapidly be overtaken. What is more, it would mean freezing the commercial deployment of AI tools that aren’t in the trial, effectively putting the brakes on innovation in the NHS rather than pressing the accelerator.

The National Screening Committee, which advises the NHS about screening programmes in Scotland, Wales and Northern Ireland as well as England, has decreed that AI tools which are not among the five chosen for the trial can continue to be used for research purposes. But local NHS Trusts won’t be able to sign commercial deals with the companies which have created them, even if they have already been tested successfully in the UK. That risks making the NHS a much less attractive environment for companies which need to recoup the large investment they have made in developing this technology.

One example - Kheiron Medical, the company which , as the BBC reported last year, successfully pinpointed 11 cases of breast cancer which doctors had been unable to spot in a trial with NHS Grampian. In a series of studies, its Mia automated breast scan technology has shown it can identify between 11 and 15% more cases than humans, potentially almost halving the workload for hard pressed radiologists. But Kheiron is facing an uncertain future in the UK. I gather that, having focused most of its activity on working with the NHS to trial and deploy Mia, the company is now concentrating on the US market.

One medic heavily involved in developing AI, though not in the breast cancer field, pointed out that Sweden was well on the way to completing the kind of breast cancer project which is just starting here. An ongoing trial conducted by Lund University involving more than 100,000 Swedish women has found in its latest phase that 29% more cancers were detected when doctors were assisted by AI than when the traditional method involving two radiologists was used.

Importantly, there was no rise in the number of false positives, where women are called back for stressful and unnecessary further tests. The results from the trial have led to several regional breast screening programmes in Sweden integrating AI into their work.

In both Sweden and the UK, the promise of AI is that it will help health services cope better with a shortage of radiologists rather than replacing them. If instead of two humans giving independent assessments of the same scan, one doctor looks at it, then compares it with the AI tool’s verdict that should boost productivity, enabling radiologists to see more patients. Sceptics will argue that over time the humans will get lazy, just assuming that the AI will have got the right answer, so the technology will need to be constantly monitored to ensure that standards are maintained.

But the concern must be that while the UK carries out what will certainly be a very thorough and authoritative piece of research, other countries like. Sweden, Germany. and the United States will be deploying AI in their health systems and improving its performance,

The AI medic who pointed me towards the Swedish study, battle-scarred after years of trying to get his ideas adopted, gave me this cynical view:

“If you're an AI company, and your business plan is to somehow make money, and your company has to live or die based on the NHS, then I think you're in a bad situation, because it's never going to happen.”