Artificial intelligence (AI) is all the rage in many fields, and ophthalmology is no exception. Symposium 13 on Day 3 of the 7th Congress of the Asia-Pacific Glaucoma Society (APGC 2024) in Manila, Philippines, focused on the latest in AI breakthroughs in glaucoma care and management.
AI in clinical practice
Integrating AI into clinical practice for glaucoma diagnosis and management presents unique challenges and opportunities. Dr. Leopold Schmetterer from the Singapore Eye Research Institute (SERI) shared his insights on this topic, exploring the potential and pitfalls of AI in everyday clinical settings.
“Increasingly, AI applications in ophthalmology are popping up in MEDLINE searches,” Dr. Schmetterer noted, highlighting the growing interest in this technology. Yet, he pointed out a significant lag in real-world clinical adoption. “When it comes to real clinical applications, we are rather at a much lower pace,” he said. This discrepancy stems from several challenges that need to be addressed.
One major hurdle is the nature of the data used to train AI models. “Many of the studies are published on pre-selected patient populations, not on real clinical applicability,” explained Dr. Schmetterer. This makes it difficult to apply these models to diverse patient groups seen in primary or secondary care settings. He emphasized the risk of “data set shifts, accidental fitting of confounders and unintended discriminatory bias.”
Logistical issues also play a significant role. Implementing AI requires sophisticated technical infrastructure and seamless integration into existing workflows, which can be daunting for many clinics. “It’s a question of cost-effectiveness that needs to be clarified,” Dr. Schmetterer added. Moreover, there’s a sociocultural aspect to consider: How will patients and healthcare providers accept AI as part of the diagnostic process?
Despite these challenges, AI holds immense promise for glaucoma care. Dr. Schmetterer outlined several key applications where AI could make a substantial impact:
- Screening Tool: AI can be a powerful tool for screening undiagnosed glaucoma cases. “There is a need for a screening tool,” Dr. Schmetterer emphasized, noting that AI could help identify cases that might otherwise go undetected. However, he cautioned about the issues of false positives and false negatives, which could overwhelm the healthcare system if not managed correctly.
- Diagnostic Support: While AI is unlikely to replace clinical diagnosis, it can certainly support it. AI can assist in analyzing optical coherence tomography (OCT) scans, intraocular pressure measurements and visual field tests. “AI for screening has nothing to do with AI for diagnosis,” Dr. Schmetterer clarified, emphasizing the different requirements for these applications. AI in diagnosis would work alongside traditional methods, enhancing their accuracy and efficiency.
- Precision Medicine: Another exciting area is precision medicine. AI can help tailor treatments to individual patients, optimizing therapeutic drug monitoring and potentially improving outcomes. “Based on a single fundus photograph, we can predict future progression with approximately 80% accuracy,” Dr. Schmetterer shared, underscoring the potential of AI to personalize glaucoma care.
Looking ahead, AI could also play a role in population health management and clinical trial design. Enhancing patient education through large language models and using AI to draft clinical guidelines are other intriguing possibilities. “Should we maybe use ChatGPT in the future for drafting such guidelines?” Dr. Schmetterer mused, opening the floor to innovative ideas.
AI-driven gait analysis
Imagine a future where a simple walk down a hospital corridor could help screen for glaucoma. Dr. Xiaofei Wang from Beihang University in China is pioneering just that with innovative AI-driven gait analysis research.
His goal is ambitious yet straightforward: to make glaucoma pre-screening more accessible, especially in regions where advanced diagnostic tools are scarce.
“Conducting pre-screening is very crucial for glaucoma,” Dr. Wang emphasized. Traditional diagnostic tools like OCT and fundus imaging aren’t always available to everyone. His approach uses AI to bridge this gap, analyzing how people walk to spot potential visual impairments.
Dr. Wang’s method is refreshingly simple yet effective. Instead of relying on expensive motion capture systems or attaching sensors to patients, his team uses everyday technology—mobile phones. “Any mobile phone that has a video recording function can perform this task,” he explained.
In their study, participants walked a six-meter path in a hospital corridor, performing three tasks: walking normally, walking while performing a mental subtraction task and walking over obstacles. The videos captured during these tasks were then analyzed using a deep learning model to extract gait parameters like walking speed, stride length and gait cycle.
Under normal walking conditions, there was no significant difference in gait parameters between glaucoma patients and those without the condition. However, when tasked with walking and performing a mental subtraction simultaneously, the differences became apparent.
The glaucoma group walked more slowly, had longer gait cycles and shorter strides. “Patients with visual field damage need great cognitive effort to maintain a normal gait and are unable to do so on certain tasks,” Dr. Wang noted.
Interestingly, when crossing obstacles, both groups showed similar walking speeds, suggesting that simple gait analysis might be more revealing under mentally challenging conditions.
The real breakthrough comes from the AI’s ability to classify patients based on their gait. By feeding the gait parameters into a machine learning model, Dr. Wang’s team could distinguish between glaucoma patients and healthy individuals. Although the sample size was small, the results were promising.
The next step? Scaling up the study to include more participants, which could solidify AI-driven gait analysis as a reliable pre-screening tool for glaucoma.
Besides Dr. Schmetterer’s insights into the integration of AI in clinical practice and Dr. Wang’s innovative AI-driven gait analysis, Symposium 13 attendees enjoyed a range of AI topics in ophthalmology, including applications of AI in angle closure, surgical navigation of ab interno trabeculotomy based on deep learning, revolutionizing glaucoma management with large language models and more.
With continued research and collaboration, these advancements could soon become invaluable tools in the fight against glaucoma, making early detection and tailored treatment more accessible than ever.
Editor’s Note: Reporting for this event took place during the 7th Congress of the Asia-Pacific Glaucoma Society (APGC 2024), held from May 24-26, 2024 in Manila, Philippines.
