#London #United Kingdom #DeepMind #AI for a Better World #Moorfields

Exploring AI's Role in Preventing Vision Impairment

In a world where over two billion individuals are affected by vision impairment, many cases of which are preventable, the role of artificial intelligence (AI) in healthcare has never been more significant. A recent episode of the original series "AI for a Better World" by Planet Classroom focuses on a powerful interview with Prof. Pearse Keane, an ophthalmologist from Moorfields Eye Hospital. This compelling discussion centers on the pressing global challenge of avoidable sight loss and how responsible AI can facilitate earlier detection, expedited triage, and ultimately, more equitable care for patients.

The Growing Demand for Eye Care

Dr. Keane reveals that in the UK, the demand for eye care has rapidly outpaced availability. Ophthalmology is now the busiest medical specialty, accounting for almost 10% of all National Health Service (NHS) appointments—around 10 million annually—a figure that has surged by more than a third over the last five years. He highlights age-related macular degeneration (AMD) as a leading cause of blindness; approximately 25% of Europeans over the age of 60 show early signs of the disease, and about 400 people in the UK develop severe forms of blindness each day. The alarming reality is that patients can lose their sight while they await necessary medical intervention.

Moving Past Hype to Effective Solutions

During the interview, host Cathy Rubin delves into the need to transition from theoretical discussions about AI to practical, measurable outcomes that enhance patient care. This includes emphasizing trust, ensuring patient consent, mitigating biases, and promoting transparency, all while placing human clinicians at the core of the healthcare system.

AI’s Triumph in Eye Scan Interpretation

A landmark study highlighted by Dr. Keane, stemming from his initial collaboration with Google DeepMind in 2018, illustrated that AI could read Optical Coherence Tomography (OCT) eye scans with accuracy comparable to leading specialists across more than 50 retinal diseases. This study, published in Nature Medicine, serves as a pivotal proof of concept that AI can facilitate comprehensive triage at scale. However, Dr. Keane notes the extensive work needed to transition from code to clinical application through rigorous regulatory processes, like those enforced by the FDA and MHRA.

Building Trust Through Data Infrastructure

Dr. Keane outlines the establishment of INSIGHT, Moorfields' eye-health data hub, which securely connects over 35 million ophthalmic images with clinical outcomes through a cloud-based system. This platform has been designed with patient privacy in mind and aims to enhance research while maintaining active public engagement. The overarching goal is to accelerate advancements in research without compromising individual security.

Maximizing Impact with Minimal Data

Another groundbreaking initiative introduced by Dr. Keane is the RETFound, an open retinal foundation AI model that was made public in 2023 and trained on around 1.6 million retinal images. By leveraging large unlabeled datasets, this model can be fine-tuned using significantly fewer labeled scans for various downstream tasks, including identifying glaucoma, AMD, and even indicators of systemic illnesses. The accompanying model code and weights are available for non-commercial research, empowering global teams to innovate responsibly as clinical applications await regulatory green-light.

Transforming Patient Care

As the standard for high-quality imaging is established in community optometry, AI decision support systems can prioritize urgent cases, thereby minimizing detrimental delays and bringing world-class expertise closer to local communities. Dr. Keane is optimistic that these advancements can foster earlier interventions and improve access equity to sight-saving care within just a few years rather than decades.

Key Takeaways from the Interview

Rubin's conversation addresses several crucial points:

• - The reasons behind Moorfields' partnership with Google DeepMind and the specific capacity challenges faced by clinicians.
• - How AI can enhance the speed of scan interpretation, efficiently triage urgent situations, and reach patients sooner while retaining human oversight.
• - The requirements for transitioning from theoretical research proofs to compliant clinical applications.
• - The role of open foundation models like RETFound in driving global innovation and trust within the healthcare sector.
• - Essential advice for young innovators on developing ethical and effective health technology solutions.

A Public Health Appeal

Cathy Rubin concludes the episode with a straightforward reminder about public health: "Early checks save sight. Please book a routine eye exam—especially if you experience eye problems, have diabetes, or are over 50." This call to action underlines the importance of proactive healthcare and the role that technology can play in enhancing patient outcomes.

About the Host and Interviewee

Cathy Rubin is a multimedia producer, interviewer, and author focused on the intersection of AI and education. She is the co-founder and CEO of Planet Classroom, where she has produced over 800 interviews concerning AI's societal impact. Prof. Pearse Keane serves as Professor of Artificial Medical Intelligence at UCL, a consultant ophthalmologist at Moorfields Eye Hospital, and the director and founder of INSIGHT, the largest ophthalmology data resource globally. His work is instrumental in utilizing AI to prevent avoidable vision loss. Tune in to the inspiring conversation on Planet Classroom's official platforms and engage with the ongoing discussion using the hashtags #AIForABetterWorld and #PlanetClassroom.