NHS IT vendors can no longer ignore technology they didn’t invent, and trusts need to integrate imaging far beyond the radiology department, if AI is to play its big role in diagnosing illnesses, writes Jane Rendall, MD, Sectra UK and Ireland.
There is no way Sectra could ever hope to develop even a fraction of the potential of artificial intelligence in diagnosing illnesses. It might sound strange for an NHS diagnostic technology provider to start a comment piece by saying what it can’t do. But this is a reality that all technology vendors face and must act on if the NHS is to realise the new ‘tech vision’ launched by the health secretary in October 2018.
The Department of Health and Social Care’s vision argues that artificial intelligence has “huge potential to improve diagnosis”. This is absolutely the case. But that potential will only be realised if another key facet of the same tech vision – interoperability – extends to the ‘ologies’, and to the imaging technology on which they rely.
Suppliers need to be open to AI algorithms from anywhere
A momentum for the NHS to move beyond outdated technology is building, and if vendors of traditional IT are to survive they must support the integration of all sorts of innovations into their offering. If they don’t do this, they will go the same way as the fax machine, and sooner than they might think.
Remarkably, whilst many companies are thinking collaboratively, we still live in an age where some diagnostic med-tech hardware suppliers force hospitals to purchase their own software to use on their modalities.
Now is the time for suppliers to be open, not archaic. Vendors can’t block the NHS from accessing innovations just because they didn’t invent it. And the same applies to innovations in the AI space.
Right now, within the NHS, and in companies across the globe, people are developing applications and algorithms to tackle real world problems – ranging from detecting deadly diseases sooner, to avoiding unnecessary appointments and encounters.
No single technology vendor could dream up even the smallest percentage of the ideas that innovators are coming up with worldwide. Yet, if you were to look across the competitive landscape in NHS diagnostic IT, you might often hear the same mantra from companies obsessed with their own algorithm developments, rather than the wealth of ingenuity they could embrace for their NHS customers.
This isn’t about what any one vendor is developing – and it is not about what Sectra is developing. The key is for suppliers of traditional technologies, like the electronic patient record (EPR) or the picture archiving and communication system (PACS), to ensure they are open and interoperable, and that AI applications demanded by their customers can plug in to the core technologies relied on in the NHS.
An open platform for AI diagnostics
Hospitals need their core systems to be open platforms for AI. Sectra is already working with hospitals to achieve this. We have a fundamental belief this is how we will be successful as a supplier.
If a hospital wishes to implement their homegrown application into their PACS to identify patients across their region at risk of cancer, they can do this.
Or if they want to implement an algorithm developed by another vendor or innovator to diagnose diabetes sooner, following necessary due diligence to ensure that application is safe, we will plug this in to the PACS.
This is our ethos – to enhance our core role around the flow of imaging and diagnoses in a way that is helpful to radiologists, pathologists or any healthcare professional – by building on this as an open platform that allows AI to securely flourish for the benefit of professionals and patients. And we will continuously work harder at this, so we can make sure we can give our NHS customers the opportunity to take advantage of the newest technologies that can make a difference.
Leaving silos behind
Really making this work however means applying AI to the richest dataset as possible. That doesn’t necessarily have to mean sharing data externally, but it certainly necessitates joining together data and imaging across the trust, and thinking at the population level.
Simply put, the NHS can’t afford to keep investing in departmental radiology PACS for hospitals if it is to expose its new AI applications to the data they need to support effective diagnoses.
Enterprise wide image management is the only way forward, where the barriers across the ologies are broken down, so that images from a whole host of diagnostic specialities can support diagnoses and become integrated with information from the EPR.
Standards such as HL7 FHIR, or Fast Health Interoperability Resources, can be key in enabling this, but again, this does require willingness from vendors to collaborate, and for the NHS to think differently about a strategy for digitising images and information.
Only then can AI be used to look for patterns in data, to identify incidental findings that might show the early onset of a disease not originally being searched for, or that can help to determine how a catastrophic event in A&E might be avoided.
These are real world problems, but only through a genuine interoperable and open approach from the NHS and its suppliers will they be solved.