AI may well be the answer... but there are three things that need fixing first

Prime Minister Theresa May delivered an important speech in May, calling for innovators and clinical staff to work closely to make the dream of AI for cancer (and other) diagnoses a full-scale reality.

It was part of the launch of a new government target to get people living healthily for five more years by the year 2035, and it incorporates the prevailing consensus in western medicine at the moment that prevention rather than intervention is the only real way to managing the ageing population. The PM also discussed other aims under the umbrella of the fourth industrial revolution.

Fittingly, the pledge also came at the same time as the UK health sector hits an important milestone - the 70th Anniversary of the NHS. 70 years is a long time in innovative terms, and at the centre of this issue you’ll find a potted history of the technological advances of the NHS through the decades - including a look at what’s to come (and yes, AI/machine learning is on the list).

In our office we have an ongoing debate about AI. Some say it’s over-hyped, and the potential is so far in the distance that it’s almost invisible. Others believe that we are just around the corner from a new age of diagnostics.

Personally, I think there are three things that need addressing - and the sooner we do, the sooner we can start predicting cancer and cardiovascular disease before they have chance to do irreversible damage.

The first is adoption. At Med-Tech Innovation Expo in Coventry earlier this year, I spoke with a number of visitors from various backgrounds who told me they still felt that government initiatives like the Accelerated Access Review are still merely paying lip service to the underlying problems of getting technology into the NHS. And, they said, despite the promising name, the Accelerated Access Review is also proving slow to deliver the goods. A common belief is that the NHS, or rather, the policymakers behind the scenes, still place too much emphasis on cost-effectiveness, rather than cost-benefit and long-term efficiencies. Implementing AI will take considerable investment - and the government has a duty to make that available if it intends to set targets around AI adoption.

Next, there’s the issue of the data itself. We know that there are already vast amounts of information available, but before it can be translated into useful algorithms, surely it needs to be organised into such a way that it can be interpreted by a machine. Much of the data we have available at the moment was never obtained with the intention that it would be used to create an algorithm. As such, there’s some good old-fashioned human work to be done there.

Which leads me on to the third hurdle that needs knocking down: skills. Thankfully, the government has placed significant emphasis on the need for more education and talent acquisition in the field of data science - part of its future-proofing strategy. There are also accelerators and incentive schemes on offer for those willing to put these skills to good use. But it’s not yet a mature talent pool, and until it is, there may not be enough pairs of hands to implement AI on the scale needed to beat cancer and heart disease.

Despite these (and probably several other) challenges, the good news is that the industry is making good progress. Just two days after the PM’s speech, I received word that Oxford University spinout Ultromics has raised £10 million in series A funding to bring its coronary artery disease diagnosis algorithm into 20 NHS hospitals. Let’s hope that the government will match the efforts of industry in making AI a reality for the NHS.