Executive Chair of InnoScot Health, Graham Watson analyses the current landscape

Artificial intelligence (AI) for improving diagnostics across the NHS is an exciting, burgeoning area – but while greater progress is undoubtedly needed, caution is still required.

We know that AI solutions have the ability to deliver benefits nationally and equitably through algorithms that analyse medical images and data – including X-rays, CT scans, MRIs, and patient records – facilitating early detection.

This ability to detect disease at an earlier point means treatments can be implemented faster, often at a point when they can be more effective. In the present, it is arguably the most promising application of AI in healthcare.

If delivered correctly, it means greater likelihood of better outcomes, increased survival rates, reduced long-term complications, and improved quality of life for patients.

Of course, a secondary benefit to early detection for a pressured NHS is lower long-term costs by reducing the need for expensive advanced treatments while realising a reduced burden on emergency services and on our dedicated healthcare professionals.

Such efficiency in diagnostics allows the health service to do more within its limited resources and focus on patient-centred care.

Speaking recently at the Scottish Parliament, Cabinet Secretary for Health and Social Care, Neil Gray insisted that “a scientific revolution is underway that has the power to transform healthcare”.

He said that this “offers genuine cause for optimism” and includes “the application of artificial intelligence to diagnose and treat disease, as well as to keep people healthier for longer”.

There are indeed AI diagnostic success stories for NHS Scotland and the figures alone are extremely encouraging. NHS Grampian’s innovative GEMINI initiative – developed with the University of Aberdeen and Cairn Technology – is assisting doctors in the detection of breast cancer by leveraging AI to improve screening accuracy.

So far, it has helped radiologists detect 10.4 per cent more cancers than standard screenings, while reducing their workload burden by 36 per cent, in turn allowing them to place a greater focus on patients rather than processes.

Only last month, the team behind GEMINI received the 'Data Driven Innovation Award' at this year's Digital Health and Care Awards, and no wonder.

It is exactly the kind of technology that ensures efficiency without compromising on quality by allowing healthcare professionals to dedicate more time to patients than to processes.

Likewise, AI-powered diagnostic successes have been reported by NHS Greater Glasgow and Clyde through the utilisation of qXR software – an AI-enhanced chest X-ray reporting solution for improved early detection of lung cancer.

That software is at the heart of the RADICAL study which is being delivered by the Living Laboratory's Digital Health Validation Lab (DHVL) alongside West of Scotland Innovation Hub.

However, while AI innovations applied to diagnostics are indeed rapidly advancing worldwide, it is widely acknowledged that Scotland is not keeping pace.

JD Blackwood, AI lead for NHS Forth Valley, recently insisted that the AI landscape in NHS Scotland has “barely changed” in the last three years despite being the first country in the UK to introduce AI in diabetes screening.

A previous guest speaker for InnoScot Health’s webinar series, Mr Blackwood, said that “to be truly successful, we must have committed leadership in government, the health sector, and social care”.

I understand his frustration at the country’s perceived stasis, but there are also complexities to account for and barriers to fully overcome as part of the bigger picture, not least through high development costs and challenges in real-world deployment.

We still stand at something of a crossroads where AI is concerned with huge opportunity having to be balanced with a safe, ethical, well-considered approach. In other words, true success and efficacy requires caution and reflection.

Indeed, the technology continues to evolve at such a pace that, through unintended AI-generated biases and even the threat of cyber-attacks, the risk of patient confidentiality breaches and clinical errors remains, so the integration of such innovation must undergo a thorough period of secure testing.

The reality is that NHS Scotland patients simply cannot be exposed to risk – whether through potential mishandling of sensitive data, perpetuating inequities in care through AI’s learning of historical data, or issues in compliance as the regulatory framework struggles to keep up.

At the very least, adaptive AI’s constant evolution requires robust, consistent, ongoing evaluation, not to mention the provision of detailed workforce training to match, helping to ensure data literacy, balanced delivery, and informed decision-making both today and tomorrow.

Data literacy in particular is a skills gap that will have to be addressed in order to critically assess results before deciding whether or not they are fit to be integrated into key decision-making processes.

Earlier this year, Neil Gray said that the Scottish Government intends to publish a framework by July “for the safe and ethical use of AI across health and social care,” which is a proactive step that is to be welcomed.

AI innovation is not just a term to be loosely coined – it must be a clear, multi-faceted commitment that is supported by decisive action, and by the workforce itself.

Staff must be assured of the benefits in order to pass trust onto the patient, and to also take the next step – of coming forward with, and then realising, their own innovative ideas that are supplemented by AI technologies.

I am in absolutely no doubt that such an AI-powered healthcare future can be hugely beneficial in supporting early interventions, saving lives, and much more in between – we certainly have the building blocks – but first, we have to be sure of getting it right in the here and now.