AI in Eye Care: Opportunities and Challenges for Aotearoa
AI in Eye Care: Opportunities and Challenges for Aotearoa
Executive Summary
Artificial intelligence (AI) is rapidly transforming ophthalmology and optometry, offering scalable solutions to address the global burden of eye diseases and the shortage of eye care professionals. AI systems are demonstrating significant potential in enhancing diagnostic precision, streamlining workflows, predicting disease progression, and expanding access to care, particularly in underserved regions like New Zealand and low- and middle-income countries (LMICs). The global AI Eye Screening System market is projected for substantial growth, from approximately US$126.95 million in 2023 to US$8,978.98 million by 2030, a CAGR of 82.55%.
However, alongside these promising advancements, critical challenges must be addressed for successful and equitable integration. These include ensuring data quality and diversity to prevent bias, developing transparent and interpretable AI models ("white box" vs. "black box" systems), establishing clear ethical and regulatory frameworks, and bridging digital divides in infrastructure and health literacy. Optometry Australia and other professional bodies emphasize that AI should complement, not replace, human clinical judgment and accountability.
Key Themes and Important Ideas
1. The Global Burden of Eye Disease and the Need for AI/Teleophthalmology
Significant Global Challenge: Over 2.2 billion people worldwide are affected by vision impairment, with nearly half of these cases preventable or treatable. This burden disproportionately affects developing nations and aging populations, with 80% of cases in vulnerable communities. The number of blind people is projected to increase to 610 million by 2050.
Workforce Shortages: There is a critical global shortage of eye care professionals. For instance, low-income countries have only 3.7 ophthalmologists per million people, compared to 76.2 in high-income countries. New Zealand has one of the lowest numbers of ophthalmologists among comparable countries (3/100,000).
Geographical and Socioeconomic Barriers: Seven hundred thousand (16%) of New Zealanders live without timely access to healthcare services. Mountainous terrain, long travel distances, and a dispersed population exacerbate these issues, particularly affecting Indigenous Māori, New Zealand Europeans, and the elderly.
COVID-19 as an Accelerator: The pandemic gained unprecedented traction for teleophthalmology, demonstrating its role in overcoming resource and distance barriers, though it also highlighted existing equity issues for disadvantaged groups.
2. Applications of AI and Teleophthalmology in Eye Care
AI and teleophthalmology are being applied across the entire patient care workflow, from screening to management:
Screening and Early Detection: Diabetic Retinopathy (DR): The most common disease addressed by teleophthalmology (17% of services identified). Programs like the UK NHS Diabetic Eye Screening Program have screened over 2 million people, leading to DR no longer being the leading cause of blindness in the UK. AI models for DR detection have achieved high sensitivity (75-91.7% for proliferative DR, 75-94.7% for non-proliferative DR). The DR segment is expected to hold the largest market share in AI eye screening systems (57.88% by 2030).
Glaucoma: AI models using fundus photos and OCT scans show high sensitivity and specificity in detecting glaucomatous changes. AI-assisted screening has reduced misdiagnosis rates by 22% in some trials and improved early diagnosis rates by 30%.
Age-Related Macular Degeneration (ARMD): AI systems can detect indicators of neovascular AMD with over 80% accuracy and specificity of 91.4% (ForeseeHome device).
Retinopathy of Prematurity (ROP): Digital retinal imaging for ROP has shown 100% sensitivity and 97.9% specificity in a New Zealand study, and AI programs have achieved 95% diagnostic accuracy. This is particularly valuable in resource-scarce settings due to lower cost and reduced stress on premature infants.
Refractive Errors (e.g., Myopia, Amblyopia, Strabismus): AI-driven photo-screeners and deep learning systems are improving early detection and prediction of myopia onset and progression in children, with accuracies up to 99.9%. AI is also being used to identify strabismus (95.2% accuracy) and amblyopia risk factors (90.8% accuracy).
School Screening: Photo-screener devices are widely used for amblyopia screening, though accuracy varies by device and criteria.
Diagnosis and Assessment: AI can provide objective, consistent, and accurate assessments for conditions like myopic maculopathy (97% accuracy), and differentiate various eye diseases from images.
Treatment and Management: Cataracts: New Zealand has the "Catrax" cloud-based service, which has cut down by 4 weeks triaging waiting times for cataract surgery by instantly assessing patient eligibility.
Contact Lenses & Spectacle Prescriptions: AI is optimising Ortho-K lens fitting and predicting lens parameters with high accuracy (e.g., 96.4–97.4% for return zone depth in Ortho-K).
Low Vision Rehabilitation: AI offers personalized solutions to improve the quality of life for visually impaired individuals, including mobile applications for object recognition (up to 99.6% accuracy) and device fitting assistance.
Referral Pathways: AI-based decision support systems can assist with referral decisions for retinal diseases, achieving 95% accuracy.
Telemedicine and Remote Care: The combination of AI and telemedicine enables remote diagnosis and monitoring of eye diseases, improving access to eye care services for people in underserved areas. This paradigm follows an asynchronous consultation system characterized by front-end portable device data acquisition and cloud-based expert diagnosis.
Live video consultations are also used in emergency eye care in Australia and the UK, reducing transfers and halving the need for second appointments.
Smartphone applications and portable devices facilitating early detection and timely intervention.
Healthcare Optimization and Education: AI aids in medical resource allocation, health education, and patient management. It can deliver tailored training for healthcare workers and simulate surgical procedures using VR/AR.
3. Market Growth and Geographic Distribution
Rapid Market Expansion: The global Artificial Intelligence Eye Screening System market is projected for explosive growth, with a CAGR of 82.55% from 2024 to 2030, reaching US$8,978.98 million.
Regional Dominance: While North America held a significant share in 2023 (35.09%), Asia-Pacific is projected to dominate by 2030 (67.44% of sales), reflecting concentrated development and deployment in these regions.
Leading Companies: Major global companies in this market include Airdoc Technology, Digital Diagnostics, Eyenuk, RetinaLyze System, Optomed, and VUNO.
New Zealand's Lag: New Zealand’s teleophthalmology services, however, are currently limited, with only three identified discrete services in a comparative review. The US leads in teleophthalmology services (54.5% of identified programs).
4. Challenges and Ethical Considerations
Despite its immense potential, several critical challenges must be addressed for the responsible and effective integration of AI in eye care:
Transparency and Explainability ("Black Box" vs. "White Box"): Many AI models are "black box" systems, meaning their inner working are not interpretable, making it difficult to understand why a particular diagnosis or recommendation was made. This hinders clinicians’ comprehension and trust.
Lack of transparency is associated with decreased accuracy of AI algorithms.
"White box" AI systems produce more linear and reliable data but are comparatively less innovative. The focus should be on creating models that are inherently interpretable or developing Explainable AI (XAI) frameworks that elucidate the rationale behind predictions.
Data Quality, Bias, and Diversity: AI models rely heavily on large, diverse datasets for training. Most datasets used in the reviewed studies lack representation across diverse demographic groups, leading to potential sample selection bias, ethnic disparities, or geographical variations. This can result in suboptimal model performance in specific populations and misdiagnose a patient based on an incomplete or inadequate training set.
Data incompleteness, inaccuracy, and bias compromise reliability and safety.
Racial bias in ophthalmologic clinical trials is an ongoing concern, and this trend could continue into AI development if it remains unchecked.
This raises concerns of health poverty if AI disproportionately harms disadvantaged groups.
Responsibility and Accountability: A responsibility gap arises when responsibility cannot be easily attributed for harms caused by AI. While some companies (e.g., IDx) accept liability for on-label use, accountability for off-label use or errors in adaptive AI models remains unclear.
Professional bodies like Optometry Australia affirm that Clinical decisions in eyecare must continue to be made by, and remain the responsibility of, health professionals who have the appropriate expertise, qualifications, and professional accountability. AI should be a decision-support tool, not an autonomous diagnostician.
Regulatory and Ethical Frameworks: Existing regulatory frameworks (e.g., FDA for medical devices) may not be well-suited for continuously learning AI systems. There is a need for dynamic certification mechanisms and stratified accountability frameworks.
Optometry Australia calls for legislated AI Principles and an ethical framework for the use of AI in health care and eye care, emphasizing informed consent, privacy, security, and clear guidance on responsible use.
Concerns exist regarding improper use of AI and AI applications that undervalue the knowledge base, communication skills, and the clinical judgement and expertise of optometrists.
Infrastructure and Implementation: Inadequate infrastructure and regulation, ethics, and sociocultural aspects hinder AI adoption.
Technical barriers include long setup periods, and difficulties with the operating systems and devices, as well as network connectivity issues.
Poor health or technology literacy and clinician fear of malpractice liability or resistance to change continue to hinder its development.
The existing model of DHB DR screening is resource intensive, requiring a team of trained clinicians to read the photographs. They also have high capital setup costs.
5. Future Directions and Recommendations
Strategic Partnerships and National Investment: Investing in strategic partnerships and technology at a national level can advance health equities in ophthalmic care in New Zealand.
Improved Infrastructure: Enhancing internet infrastructure (e.g., community-led internet service subsidies, WIFI-hotspot programs) and technology distribution (e.g., smartphones with community training) is crucial.
Standardization: Developing standardized data collection protocols and interoperable platforms for data sharing and training AI models.
Human-Centred Design: Involving clinicians and patients in the design and development process of AI systems, and conducting real-life evaluations in the clinic.
Continuous Learning and Education: Optometrists and other healthcare professionals must stay abreast of the latest developments in AI and receive continuing professional development (CPD) on AI technologies, data interpretation, and ethical use. Optometry schools should integrate AI into their curricula.
Focus on Equity: Prioritizing AI applications that improve access and equity, especially for disadvantaged communities, without compromising quality of care. Avoiding AI that could widen the access gap.
Regulatory Evolution: Developing dynamic regulatory frameworks that allow for continuous learning in AI while ensuring safety, efficacy, and accountability.
Conclusion
While AI offers a powerful tool to address significant challenges in eye care, its successful integration requires a careful and coordinated approach that prioritizes patient safety, ethical practice, and equitable access. New Zealand, in particular, has a significant opportunity to invest strategically in teleophthalmology and AI to improve access for its underserved populations.
