- Tomorrow's Dose
- Posts
- Edition 13 - AI in Medicine: NHS Adoption, Postmarket, and UK Strategy
Edition 13 - AI in Medicine: NHS Adoption, Postmarket, and UK Strategy
Explore AI’s adoption by the NHS, its role in postmarket monitoring, and the UK’s balanced approach.
Royal College of Radiologists urges NHS to embrace AI to cope with demand
Methods and tools for effective postmarket monitoring of AI-enabled medical devices.
AI in medicine: UK is pursuing 'middle path' in adoption and regulation
Featured follow of the week
Top posts of the week across social
Meet the editor
Want a featured article?
This week’s edition is based on the recent Royal College of Radiologists Global AI Conference on 3-4 February 2025 - LINK.
Presenting our work on AI autocontouring, which eventually won the People’s Choice award VIEW, I was fortunate to attend the full programme of excellent talks. This newsletter highlights some of the important points of the conference.
Specialty: Radiology // Sub-Specialty: AI // Body Site: All
1. Royal College of Radiologists urges NHS to embrace AI to cope with demand
Speaking at the inaugral Royal College of Radiologists Global AI conference in London, the RCR president Dr Katherine Halliday said the UK should 'urgently embrace' the rollout of AI, which could boost productivity, speed up diagnoses, and free up doctors’ time for patient care. The RCR was said to be 'committed to harnessing the potential of AI while mitigating the risks', but also stressed the need for an increase in the number of doctors. There is currently a 30% shortfall in the number of consultant radiologists, which could increase to 40% by 2028.
Read Full Article
Paul’s Thoughts:
It is great to see the RCR endorsing and embracing AI tools, which have a huge potential to reduce the burden on struggling, overwhelmed, radiology departments in the NHS. In the RCR's 2023 workforce census it was found that 54% of all radiology departments already use AI tools, and the number is expected to have grown since then. Despite all the benefits of AI, the severe workforce shortage (currently at 30%, equating to 1962 doctors) is the critical factor behind the persistent failure to meet cancer waiting time targets. It was discussed at the RCR meeting that making AI a fundamental component of radiology care could improve this number, not as a replacement for the doctor but to make the field more attractive to tech-savvy graduates. Of 1016 FDA-approved medical devices with AI and machine learning, 777 (76%) are in the field of radiology - it is clear there is great potential with a range of options available.
Timescale: Acute | 0 Years
Specialty: All // Sub-Specialty: AI // Body Site: All
2. Methods and tools for effective postmarket monitoring of AI-enabled medical devices
The FDA's Center for Devices and Radiological Health (CDRH) is actively working on research to enhance postmarket monitoring of AI-enabled medical devices. This initiative focuses on creating methods and tools to detect input data changes, monitor output performance, and understand causes of performance variations, ensuring the safety and effectiveness of these devices in clinical settings. Key projects include detecting out-of-distribution inputs, proactive data drift monitoring, and real-world AI model evaluation through federated frameworks.
Read Full Article
Paul’s Thoughts:
As a medical physicist responsible for post-deployment surveillance, I find the FDA’s initiative on AI-enabled medical device monitoring both timely and essential. AI models can degrade over time due to data drift or unexpected clinical variations, making real-time performance tracking a critical safeguard for patient safety. The emphasis on out-of-distribution detection is promising, as it enables proactive risk mitigation without compromising patient privacy. This was a topic with a lot of discussion during the RCR Global AI Conference, with questions about how to monitor AI that is not a ‘frozen model’ and updates over time. The FDA has plans to introduce pre-determined change control plans (PDCCP) being necessary for algorithms that update over time. But the FDA is not prepared for AI that updates itself - this is the next frontier that we can expect in healthcare and it presents a real headache for post-deployment surveillance.
Timescale: Acute | 1 Year
Specialty: Radiology // Sub-Specialty: AI // Body Site: All
3. AI in medicine: UK is pursuing 'middle path' in adoption and regulation
Dominic Cushnan, deputy director for AI at NHS England, gave the opening keynote speech at the RCR Global AI Conference in London on 3 February and said the UK was 'creating a regulatory environment that both protects patients and encourages innovation by providing this balanced framework'. With the less regulated FDA and strictly regulated EU (under the EU AI Act), it was suggested that the UK could adopt a middle ground approach. The theory being that the UK would act as a fertile testing ground for AI developers, but in a responsible manner.
Read Full Article
Paul’s Thoughts:
Discussing at the RCR Global AI Conference, Cushnan discussed the CERSI AI - Centre of Excellence for Regulatory Science and Innovation in AI and Digital Health (https://www.cersi-ai.org/). Through this framework, the aim is to work with innovators, regulators, patients and the NHS to optiimise the regulation of AI, to accelerate innovation. The aim is for the UK to sit as the middle ground between the US (under the FDA), which has relatively lax laws for approval, and the EU (under the EU AI Act), which has rather strict constraints for approval. The UK aims to develop itself as a global hub for responsible AI development and deployment. Also mentioned was the MHRA AI Airlock programme (https://www.gov.uk/government/publications/ai-airlock-pilot-cohort/ai-airlock-pilot-cohort#overview) - a sandbox for proactive, collaborative, agile testing of AI as a Medical Device (AIaMD), but in a safe manner. It is hoped that these initiatives show the UK is leading in terms of safe, responsible deployment, but in a manner that is not restrictive to innovation. It remains to be seen if this approach will be effective - we will probably know more in 1 or 2 years' time, given the rapid speed of progression in the field.
Timescale: Acute | 1 Year
A round-up of some of the best posts we found online this week.
Was this email forwarded to you?
Our weekly email brings you the latest health trends and insights, combining top news and opinions into a straightforward, digestible format.
 |  |  |
Want an article featured?
Have an insightful link or story about the future of medical health? Reach out below, and we may include it in a future release.
Reply