How do you use technology to transform healthcare and positively impact patient outcomes? It’s a question with scores of possible answers – but it’s one that Arm, Great Ormond Street Hospital (GOSH) and partners are trying to address as part of GOSH’s Digital Research, Informatics and Virtual Environments (DRIVE) initiative, which launched today.
The goal of the initiative is to harness innovative technologies and rapidly embed them in hospital practice – first within GOSH itself, and then throughout the NHS. Within this partnership, Arm’s focus is on artificial intelligence (AI) at the edge – that is, on the devices that we hold in our hands or employ every day in our homes and workplaces. Each of these devices has the potential to capture and process data in real time, providing instantaneous situational analysis, identifying patterns and prompting swift, decisive action to enable better experiences or avert problematic scenarios.
The exciting thing about collaborating with GOSH is that the possibilities are literally infinite – and each of those possibilities has the potential to improve a sick child’s life.
Costs are always an issue for healthcare providers so, as we tried to narrow down our initial ideas, we began looking at things at a macro level – for example, how can we streamline processes, and help GOSH to deliver cost-effective care? But we were also interested in looking at things from a patient’s perspective.
AI has the versatility to be applied to many different scenarios, so we were confident it could provide a compelling response to a range of issues in the healthcare space.
AI is already having a significant impact across virtually every market: it’s revolutionizing the way we interact with our devices; it’s pushing the dramatic advances in self-driving cars; and it’s also affecting less obvious sectors, such as farming and logistics. The potential for AI is so far-reaching, it’s hard to imagine a sector that won’t be affected – and healthcare is no exception.
News headlines are full of examples of fantastic AI applications that are having a dramatic impact on patients’ lives: a team of experts from Moorfields Eye Hospital and Google’s DeepMind recently created an algorithm that enables computers to analyze high-resolution 3D scans of the back of the eye and detect more than 50 eye conditions. Amiko’s Respiro, the smart asthma inhaler add-on, tracks and reports on inhalation technique, flow rate and volume, providing insights to the patient’s care team and shaping future treatment plans. There are apps in development that aim to analyse suspicious-looking moles and alert the user if they appear cancerous.
But while these ground-breaking solutions are undeniably making great strides to advance patient care, in talking to GOSH, we discovered that many of the AI workloads that bring the most operational benefits are far more prosaic. There’s significant value in optimizing the ‘hidden’ elements of healthcare, which many not be headline-grabbing but can still have a huge impact. These are not necessarily on the frontline, but by streamlining actions that are performed hundreds or thousands of times a day by doctors, surgeons, accountants, secretaries, receptionists – actions that oil the wheels of each and every healthcare institution – you can not only make significant savings in time and money that free up resources to be used at other points of the care chain, you can also ensure consistency of service and provide additional safeguards in, for instance, operating theatres and surgeries.
With that in mind, we decided to look at the patient experience right from the start of their journey, from the moment a child and their family walk through the doors and into the reception area.
Hospital receptions are notoriously busy places. At GOSH, at any given point of the day, there may be patients and their families, doctors, nurses, ambulance crews, security staff, cleaning staff, visitors… With so many people coming and going, ensuring that staff, patients and visitors are in the right place at the right time is something the hospital has to take seriously – particularly given the tender age of their patients. Adding AI to their existing procedures could promote efficiency, reduce admin time and provide an extra layer of assurance.
After careful consultation, we devised an AI-based person and object recognition system – based on the first iteration of the Arm Object Detection processor – to be trialled in GOSH’s new DRIVE unit. This would make registration of visitors quicker and easier, grant or restrict access to zones of the hospital, and track the location of key staff members – all based on the individual’s identity. Any data generated by the system is processed on the devices themselves, for a number of compelling reasons…
As anyone who’s waited impatiently for a web page to load knows, sending data back and forth to the cloud can result in latency – something that’s annoying for a consumer, but which time-critical applications simply cannot permit. By contrast, AI at the edge brings increased responsiveness, driving the immediate insights and prompt actions that can be vital in a healthcare setting. But keeping data on-device can also bring cost benefits: storing and repeatedly moving data is expensive, consuming budget that can be better deployed elsewhere. Furthermore – and perhaps most important in this context – repeated data transmission can have implications for the security and privacy of sensitive information: the more data is moved around, the more opportunities there are for its integrity to be compromised. By keeping as much as possible on-device, access is restricted and risk is reduced.
Operated via the receptionist’s app, our system, which is currently being developed in the DRIVE unit, provides automatic check-in of known visitors – those already in the database, such as permanent staff – and recognition of new visitors. Entrance and exit times are automatically logged, to give a clear picture of who is in the building at any one time.
The location of individuals is monitored – making it easy to track staff who may be required urgently – and the receptionist can update access permissions, via the app, in real time. Unauthorized access is quickly flagged and time stamped, with the option to take action or simply grant access. Edge compute means that analysis of data is much faster: in this real-time situation, latency would not be acceptable.
The beauty of the system is that it’s invisible, “frictionless” technology: neither visitors nor staff are required to do anything. They don’t need to scan their irises or have their fingerprints read – they simply walk through the door. And because the technology uses metadata, rather than full streaming video, individual identities are secure. This economy of data also means that the system can support a near-unlimited number of cameras, in any configuration, making it easy to port across multiple sites, reducing overall cost of operation.
It’s clear that AI will be transformative for healthcare, but we need to go beyond the high-profile, headline-grabbing use cases to find practical solutions that will streamline processes for hospitals, and make life easier for patients.
We’re just at the start of our journey with GOSH; there’s so much potential and we’re only beginning to scratch the surface. This ongoing collaboration is allowing us to understand the real-world needs of hospitals, patients and staff and turn raw technology into invisible and invaluable solutions.
DRIVE’s remit is exciting and ambitious, and promises benefits that will extend far beyond the UK health system. We have the opportunity to revolutionize healthcare as we know it. Let’s see how far we can go.
For more information about machine learning on edge devices read our machine learning solutions brief.
Edge Intelligence will become increasingly prominent as the data generated by IoT (and especially video) enables insight for competitive advantage. I read recently that by 2021, more than 80% of all IP data will be video and currently only 5 - 10% of it is analysed. Says it all.