HOT TOPICS IN OPHTHALMOLOGY INNOVATION:
ARTIFICIAL INTELLIGENCE
With all the recent buzz about artificial intelligence (AI) in medicine, you might be wondering how it might be used to support eye care. Dr. Fares Antaki shares his perspectives on the potential role of AI technology in the field of ophthalmology.
Dr. Fares Antaki
Ophthalmologist
Centre hospitalier de l’Université de Montréal,
Montréal, Québec
Why does ophthalmology need AI?
More patients
An aging population means more patients who require eye care
Fewer ophthalmologists
The number of ophthalmologists in Canada isn’t expected to keep up with the growing patient load
Limited access to care
Patients in remote or rural areas face barriers to accessing eye care
“I believe we need to build AI systems that are useful for the clinician, for the patient, and for the healthcare system. It’s important for us to determine what the bottlenecks are for providing efficient and high-quality care for our patients and using AI to address those issues.”
– Dr. Fares Antaki
Tapping into AI to positively impact eye care in Canada
While still in its infancy, AI has the potential to be integrated into ophthalmologic clinical practice with the goal of supporting three main areas:
PATIENT SCREENING
Provide large-scale, low-cost screening for some Big 5 eye diseases, like diabetic retinopathy (DR) and in the future, AMD and glaucoma
AI-assisted screening may help:
- Allow more patients to be assessed for eye disease
- Speed earlier detection of serious eye conditions
- Reduce workload on busy clinics and hospitals, allowing more time for patients who require treatment
Eventually, there’s a potential to reinvent eye exams using AI-based tools at every step to:
- Take patient history
- Measure visual acuity
- Identify eye disease through automated slit lamps and imaging
This could transform the clinic waiting room into a pre-testing space or give ophthalmologists the ability to take care of patients in remote areas
DIAGNOSTICS
Assist eye care professionals in making a diagnosis or predicting risk of disease progression with better accuracy
AI models are in development that can help:
- Diagnose diseases like glaucoma, AMD, and DR from retinal fundus images
- Identify and follow conditions of the front of the eye such as cataracts, keratitis, and conjunctival conditions
Optometry clinics
could employ AI models that help detect signs of eye disease and triage patients for referral to their nearest ophthalmologist based on the urgency
Ophthalmology clinics
could have more advanced clinical decision-making tools (for example, an AI system that considers patient demographics, optic nerve appearance, and visual fields to inform risk of glaucoma or risk of progression)
TREATMENT
Advance precision medicine – choosing the best treatment for each unique patient
Cataract surgery
AI algorithms already allow ophthalmologists to select the best power of intraocular lens for each patient based on their eye dimensions
AMD/DR
A smart AI agent is in development that informs the specialist performing anti-VEGF injections whether they need to switch medications or adjust the treatment interval, based on the appearance on imaging and patient’s visual acuity
Retinal surgery
In the future, AI systems could allow ophthalmologists to determine the ideal surgical technique and post-operative instructions for each patient for the best possible outcome
“There are many ways to fix a retinal detachment – which technique is right for this patient?”
High tech gizmos and gadgets on the horizon
Visual assessment devices equipped with a virtual assistant that can coach patients through a visual field assessment without a technician
RETFound – the first ophthalmic foundation model
- A foundation model for retinal images
- It is trained (through self-supervised learning) on over 1.6 million unlabelled optical coherence tomography (OCT) and colour fundus photography images
- The model can then be fine-tuned for a variety of disease-related clinical tasks such as:
- Detecting or classifying eye diseases such as DR, glaucoma, and AMD
- Predicting AMD progression
- Predicting occurrence of cardiovascular or neurodegenerative diseases based on retinal images
Foundation models are a novel paradigm for building AI systems. They can be multimodal – so they can understand different types of data, including text and images. Since ophthalmology is a speciality that is multimodal (relying on imaging from different machines to make a diagnosis), foundation models have a benefit over traditional deep learning models that can only take one modality at a time.
- Large-language models (text-based)
- They have shown impressive capabilities in answering questions related to ophthalmology
- Many potential applications are possible, including helping clinicians:
- Triage patients
- Assist in the diagnosis of eye diseases
- Assist in clinical documentation
Bringing an AI-based care model into routine clinical practice
While AI has the potential to transform many areas of eye care, there are several barriers to widespread adoption in Canada. Rolling out an effective AI system requires:
Electronic health records
(many provinces don’t have them, or they aren’t shared across healthcare settings)
IT infrastructure to support sophisticated AI technology & tools
Clinical guidance and strong frameworks to support ophthalmologists so they understand how to use AI tools effectively and responsibly
Even once these hurdles are addressed, AI won’t be adopted overnight. There are important questions to answer and steps to take before AI technology can be put into practice.
The pathway toward implementing an AI-based care model
IS THE MODEL EFFECTIVE AND RELIABLE?
Find an AI model with
good performance
metrics
DOES IT WORK IN OUR PATIENTS?
Validate the model in
the local patient
population
HOW DOES IT FIT INTO THE CURRENT CARE MODEL?
Integrate the model
into the current care
pathway
IS IT COST-EFFECTIVE?
Ensure the new
pathway is in line
with the cost of the
current standard of
care, while staying
safe and effective
for patients
HOW DO WE MAXIMIZE OUR CHANCE OF SUCCESS?
Put the right
mechanisms in place
to maintain patient
data confidentiality,
perform audits to
maintain
performance of the
system, etc.
Balancing the art and science of medicine
There’s certainly a place for AI in the future of eye care. AI offers opportunities to complement and support the expertise of ophthalmologists, ultimately leading to delivery of better care.
“As AI will continue to get better; we have to invest in two things: surgery and compassion. I don’t see any automated tool replacing ophthalmologists performing any type of surgery anytime soon. And I don’t think AI is here to make medicine inhumane, on the contrary. It will give us time to spend with patients that really need us, but for us to be good at that we need to invest in compassion and in our ability to care for patients as a whole.”
– Dr. Fares Antaki
