This blog is the first in a three-part series exploring the key challenges that NHS teams are facing when it comes to deploying AI. In this first instalment, we explore some of the decisions that trusts and imaging networks must consider when developing an AI strategy that is fit for the future.

Any strategy relies on funding and resources to be able to deliver it and AI projects are no different. The funding released by the Department for Health & Social Care was intended to accelerate AI adoption, but the short timescales put pressure on trusts to hastily devise strategies, against a backdrop of an ever-evolving technology landscape. So how can healthcare organisations ensure they make the best use of available resources to maximise localised innovation?

The pragmatic potential of AI in healthcare

Dominic Cushnan, Director of AI, Imaging and Deployment at NHS England is regularly quoted promoting a pragmatic view of AI’s role in healthcare. In a recent interview with Diginomica, Dominic said: “Artificial intelligence has the potential to transform healthcare delivery, both for the clinician, and for patient outcomes. But the role of artificial intelligence on the clinical side is very much to augment, and not to replace, human expertise. We need to be clear that these types of technologies are there to support clinicians in making their decisions.”

When developing an AI strategy, it is important to not only consider use cases which are developing now, but ensure you are positioned to be able to maximise the opportunity of emerging future uses within your organisation. Depending on your organisation’s digital maturity level, common use cases for AI in healthcare include:

• Disease diagnosis and prediction
• Personalised medicine
• Robotic surgery and automation
• Clinical decision support
• Remote patient monitoring
• Administration and worklist prioritisation
• Waiting list validation
• Genomics and precision medicine
• Virtual health assistants

Some of these use cases, such as disease diagnosis through the analysing of medical images, are more mature than others. However, when developing a strategy, it is important to consider the broader, evolving picture and not lock yourself out of future innovation.

Learning from the mistakes of the past

The NHS has long struggled with significant interoperability challenges partly due to vendor lock-in, which has restricted the seamless sharing and integration of healthcare data. This is a result of proprietary systems that are not designed to communicate with external platforms, creating data silos.

Arguably AI is the most complex and rapidly evolving digital specialism that healthcare teams will embrace. But AI in healthcare is at an inflection point. The majority of the deployable innovation is currently focused on radiology workflows, and PACS and RIS suppliers are hastily developing their own costly proprietary AI add-ons.

At the other end of the spectrum is an emerging ecosystem of innovative AI model creators each trying to capture their own corner of the market with bespoke point-to-point integrations. This isn’t a sustainable approach due to the lack of integration engineers, a reliance on supplier roadmaps and costly maintenance overheads.

If the current state of legacy technology solutions in the NHS has taught us anything, it is that both of these approaches create bottlenecks and stifle innovation.

Choosing the right strategic foundations

NHS digital teams have a decision to make with regards to how they will fundamentally deploy emerging AI technologies in the next three to five years:

1. Embrace their deployed PACS bolt-on offering
2. Undertake bespoke point-to-point integrations for each individual AI model selected
3. Implement an open deployment platform approach

AI is moving fast. The only option to ensure you are not locked out of future innovation is an open platform approach.

What else must digital teams consider?

Every organisation will have a different strategy depending on what their priority use cases for AI are. However, there are some common aspects that should be considered when defining a strategy.

Evaluation

AI is an emerging technology for healthcare. This means new systems deployed into clinical workflows must be thoroughly evaluated before they are used in direct patient care.

An open AI deployment platform approach enables an organisation to run multiple AI models in shadow mode before moving to full commission. Each model can first be tested on fictional data, followed by retrospective data before moving to a live clinical environment. With a multi-vendor P2P integration approach evaluation is more time consuming due to the upfront investment in resources.

Technology

A key technology infrastructure consideration of any AI strategy is whether to deploy a cloud-based AI solution, or on premise. Following a public cloud-first approach aligns with the NHS’ architecture principles. In 2017, Guy’s and St Thomas’ NHS FT was the first trust in the country to use a PACS in the cloud and since then we’ve seen a steady trend moving towards cloud-native enterprise imaging systems for the increased security and ease of access to data that they enable.

Embracing a cloud-first approach facilitates scalability and brings trusts in line with the preferred approach of many AI applications. Thereby providing access to the most current AI applications, offering clinicians cutting-edge tools to enhance patient care.

While this is an obvious choice for trusts that have already adopted a cloud strategy and have their PACS and RIS systems hosted in the cloud, the truth is that many trusts have not yet taken this step. For these trusts, deploying an on-premise pseudonymisation service instils confidence by ensuring that patient data remains protected within the trust's boundaries. This hybrid approach still allows them to tap into the full range of AI solutions available in the market whilst protecting patient data.

People and operational

Deploying AI solutions into clinical workflows is a significant challenge for both the digital teams responsible for managing and maintaining deployed solutions, and for the clinical and operational teams who evaluate them before building them into clinical workflows. Following an open deployment platform approach means that digital teams only have one new platform to learn, rather than an unlimited number of individual models that could be tested and deployed. For clinical evaluators, a platform approach enables evaluations to be carried out in a consistent manner, ensuring the most appropriate solutions are chosen to be deployed for direct patient care.

Security and information governance (IG) teams play a pivotal role in streamlining the integration of new AI applications. This is enhanced by embracing a platform and the consistent and scalable approach to process and governance that comes through familiarity with the technology.

Additionally, PACS managers benefit from this approach as they only need to establish connections to the platform rather than setting up individual point-to-point connections with each application.

Post-market surveillance and assurance

While the potential of AI is great, it also poses risks if it is not managed properly. AI devices have unique attributes that require more proactive surveillance than other software medical devices (SaMD) to ensure they are not evolving over time in a way that impacts patient safety.

A platform approach to the deployment of AI applications enables a holistic, consistent and scalable monitoring solution for the variety of applications deployed. This strongly supports continued innovation by providing the ongoing clinical assurance that models are safe and scalable.

While the statutory onus is on AI model vendors to ensure the efficacy of their models, having a platform that continually monitors deployed AI in near real-time puts the power back in the NHS trust’s hands.

By monitoring AI models as they evolve over time, using a platform approach makes it easier to check they’re performing as they should, while defending against unintended bias.

How can Answer Digital help you?

AI is moving fast. The only option to ensure you are not locked out of future innovation is an open platform approach. Get in touch with us today to make this a reality in your NHS trust.

If you’re a trust or imaging network that is looking to build your AI strategy, source funding for your AI strategy, or looking to deliver your funded strategy to meet national targets, we can help.

You can read more about our AI deployment services methodology on our website, or get in touch with our friendly expert AI team today.