Enabling The Development And Application Of Artificial Intelligence In The NHS

THE PROMISE OF AI IN HEALTHCARE

Artificial intelligence (AI) is a field of study which tries to get computers to behave in ways we would consider intelligent if those same behaviours were exhibited by humans - for example, the replication of human cognitive skills such as problem solving. But AI has huge potential beyond the mimicking of human behaviours - it is a fundamental technology that can allow meaningful processing of data beyond the comprehension of the human brain.

The promise of AI for healthcare is thus clear - it could allow every diagnosis and treatment to be personalized on the basis of all known information about a patient, incorporating lessons from collective experience. By streamlining workflows, providing automated diagnosis of routine, non-serious conditions, and by allowing liberation from keyboards, AI could ultimately also provide healthcare professionals the "gift of time" - the dedicated time really required to provide the best possible care for patients.

OBJECTIVE

Much of the recent progress in the application of AI to healthcare has come in the evaluation of eye disease. My fundamental vision for this fellowship will be to drive the development and application of AI-enabled healthcare, both in the NHS and globally, using ophthalmology as an exemplar for other medical specialties.

TRAINING AND DEVELOPMENT

Renewal of this fellowship will allow me to greatly enhance my standing as a leader in clinical AI, consolidating the leadership and technical skills I have developed to lead a multi-disciplinary research group, establish international networks, and drive innovation. CASE FOR SUPPORT

FLF renewal will support a portfolio of interlinked research projects that cover the broad spectrum of clinical AI, going "from idea to algorithm" and "from code to clinic".

A central focus of my team's early stage work will be on the scaling and validation of a foundation model ("RETFound'') that we have recently developed for ophthalmology. By going from 2 million to 20 million images in training, we will create a model which can be used in less common retinal diseases and which performs well across different demographic groups.

My team will also use AI to learn more about the most common sight-threatening retinal diseases, such as age-related macular degeneration (AMD) and diabetic retinopathy. We will develop systems that can predict disease progression, treatment burden, and visual outcomes, allowing better treatment and reducing sight loss. We will also continue to explore the emerging field of "oculomics" - using AI in an attempt to predict the future development of systemic diseases such as Alzheimer's, stroke, and heart attack.

In parallel, my team will explore the clinical validation and translation of the most promising AI systems identified from our early stage exploratory work. This will involve evaluations of diagnostic accuracy and clinical safety, closely linked with requirements for regulatory approval and subsequent health services delivery.

POTENTIAL APPLICATIONS AND BENEFITS

Renewal of this fellowship will benefit the research community through the further development of AI systems in ophthalmology, making them open source or working with industry partners to explore commercialisation as appropriate. In tandem, renewal will allow development of new approaches to validation, both in-silico and in clinical studies. Through the creation of benchmark datasets, it will assist regulators to ensure the safety and effectiveness of AI systems before they are implemented in the real world.

Most importantly, these systems will ultimately provide direct benefits for patients with better diagnosis, treatment, and monitoring of eye disease, as well as potential screening for systemic disease. Finally, the NHS will benefit by reducing pressures on already over-stretched hospital eye services, reducing the risk of patients losing vision unnecessarily.