Over the recent years, artificial intelligence (AI) has gained significant attention in various fields. Starting from its significant presence in healthcare, it has also revolutionized sectors like finance, manufacturing, customer service, marketing, and education.
AI is not a luxury post-pandemic. Despite such assumptions by few. Its impact transformed various fields as such much before 2019. The term itself was coined in 1956.1 Its application in medicine began with diagnosing using ranking algorithms, followed by its application in suggesting antibiotic treatments for potential pathogens.2
In terms of eyecare, AI has played a significant role in areas like interpretation of visual field reports, fundus photographs, diabetic retinopathy screening, and much more.
What is artificial intelligence?
In simplest terms, AI is a collective term for computer technology that attempts to replicate human intelligence. When technology does what normal humans do with “intelligence,” we call it AI.
While it may sound simple, its underlying concepts are as complex as human cognition. Some of its core concepts are machine learning, deep learning, neural networks, natural language processing, etc.3
Table 1: Definitions and examples of core concepts for AI.
| Core Concept | How Does It Work? | General Examples | Possible Applications in Eyecare |
|---|---|---|---|
| Machine learning | Algorithms learn from data and improve its performance over time without being programmed explicitly. | Netflix: Suggestions of movies/series based on your history of views. | Distinguishing glaucomatous nerve fiber layers from normal on OCT scans. |
| Neural networks | Networks of algorithms interact to process and analyze data just like neurons do in the human brain. | Facial recognition in smartphones. | Monitoring disease progression with fundus photos / OCT scans. |
| Deep learning | Uses complex neural networks in layers for more advanced processing. | Self-driving cars like Tesla. | Image segmentation to detect diseases based on an OCT / fundus photo. |
| Natural language processing | Programs devices to understand natural language, enabling easier interactions with humans. | Virtual assistants like SIRI / ALEXA that respond to voice commands. | AI algorithms for medical transcription and electronic health record analysis |
Now, let us explore its applications in the field of eyecare. In this article, we will review the current, successful integrations of AI into primary eyecare and also the other recommendations for its applications for the betterment of patient outcomes and clinical efficiency.
Current applications of AI in eyecare: A brief overview
In eyecare, the potential of AI integration is immense, given that the diagnosis and monitoring of most conditions mainly depend on imaging such as fundus photography, optical coherence tomography (OCT), and slit lamp photos.
AI tools based on deep learning and neural network algorithms have already been applied and also shown remarkable success in diagnosing, monitoring, and managing various ocular and systemic conditions. Significant applications include detection of anterior and posterior segment diseases as well as diagnosing systemic diseases based on retinal images.
In diseases like diabetic retinopathy, glaucoma, and age-related macular degeneration (AMD), the prognosis depends on early detection and progression prediction. This is where AI has proven to be efficient.
Here are a few successful applications of AI in various modes for specific posterior segment diseases:
- Diabetic retinopathy: Automated systems to detect changes in the fundus, deep-learning-assisted early detection, and progression prediction for DR.
- Two FDA-cleared AI systems for detecting DR include the IDx-DR screening system and the EyeArt system with the point-of-care screening. In a typical clinical sitting, it is difficult to detect DR with just the fundus images. But these AI systems have proven efficient in increasing the efficiency of detection.4,5
- Glaucoma: Screening with deep learning algorithms in fundus and OCT imaging. Techniques, including Visual Field Archetypal Analysis (VF-AA) and Vision Transformers (ViTs), are advancing glaucoma diagnosis by identifying patterns in visual field defects and analyzing fundus photography images.
- Large language models (LLMs) like ChatGPT and GPT-4 have also shown potential in diagnosing glaucoma and triaging and enhancing glaucoma care.6
- AMD: Prediction of AMD stages using fundus and clinical data. AI models, particularly convolutional neural networks, are capable of automated detection, diagnosis, and staging of AMD using the OCT images and fundus photographs.7
When it comes to anterior segment diseases, like cataracts, AI-powered platforms are used for surgical assistance to ensure higher accuracy and precision. Hybrid models and neural networks are used to predict keratoconus progression using tomographic inputs and for classification of subclinical keratoconus.8
Even in conditions like meibomian gland dysfunction (MGD), a multi-layer deep convolutional neural network is shown to effectively distinguish obstructive and atrophic MGD using laser confocal images.9
From detecting eye diseases with fundus imaging, AI is also being expanded to predict systemic diseases based on their retinal manifestations. Cardiovascular diseases, chronic kidney diseases, and even Alzheimer’s disease are known to have ophthalmic signs in the form of retinal manifestations. These include vascular changes like arteriolar narrowing, hemorrhages, and amyloid deposits.
Ways AI can be utilized in primary eyecare
Vision impairments due to refractive error continue to be a major concern. The economic burden due to uncorrected refractive error is estimated to be $244 billion, according to the World Health Organization (WHO).10 Thus, more screening programs, management strategies, and advanced interventions are being implemented towards refractive correction.
Refractive correction should be focused on in primary eyecare facilities as they are more accessible to community members, especially in rural areas. The integration of AI into primary eyecare can help provide more effective services.
AI-assisted refractive correction has been proven to be more precise, as it provides data-driven solutions and tailored treatments. Primary eyecare services can implement these solutions to enhance their functioning.
Here are a few suggestions for AI incorporation in primary eyecare to increase effectiveness:
- Eyeglass and contact lens prescription
- AI-enhanced refractometers are proven to provide quicker and more precise results, improving accessibility for underserved populations and busy clinical settings.11 Automated calculations and technologies for lens selection based on visual needs can streamline processes and improve patient satisfaction with customized solutions.
- Contact lenses are being well utilized for customized care. In addition to these , AI integration into the lenses can also analyze factors like visual needs, tear chemistry, and aberrations and recommend ideal power, type and design.
- Smart contact lenses are also capable of measuring intraocular pressure (IOP), glucose levels, and other factors via AI-integrated biomaterials, biosensors, and microfluidics of tear film.12
- Usually believed to be possible only in secondary or tertiary services, with AI reducing manpower in other areas, more personalized contact lens services for both refractive correction and systemic health monitoring can be recommended, even in primary centers.
- Screening patients for vision changes
- AI-powered tools such as automated screening devices and smartphone-based applications can be of huge help in analyzing real-time data, detecting minute vision changes, and providing appropriate and timely referrals to secondary or tertiary centers. This can aid early diagnosis and management of certain diseases.
- Predicting myopia progression
- An increase in myopia and myopia progression is also a major concern in terms of visual impairment. Of late, developments in the form of specialized contact lenses, orthokeratology and other technologies are being implemented to address this concern.
- AI can ease these implementations. Deep-learning techniques based on corneal topography assist in orthokeratology treatment. They can also aid in myopia screening and progression prediction.
- Existing tools include health applications and web-based tools that enable remote, continuous disease monitoring, such as SVOne, a portable wavefront aberrometer that attaches to smartphones, providing objective refractive error measurements similar to other methods.13
- AI algorithms to analyze axial length data, refractive error, lifestyle practice can also be introduced in primary eyecare to track and monitor myopia progression. Smart bands with sensors and AI algorithms can be recommended to ensure healthy visual practices by providing triggers in case of long-term near work, prompting the users for outdoor exposures and thereby, in a way, control the progression.
- Administrative streamlining and patient enhancement
- In addition to the clinical aspect, AI can also be easily leveraged to streamline administrative support. Primary eyecare services can utilize these tools to reduce the dependence on administrative support staff.
- AI-automated software can be used for appointment scheduling, billing, and much more. This eventually frees up healthcare professionals to focus on direct patient care.
- While maintaining ethical standards, AI can reduce the manual health record management. Programming the software with AI-powered algorithms to ensure accurate and error-free documentation can further enhance patient care. Natural language processing models can be used as scribes in busy clinical settings.
AI usage to improve eyecare for the underserved
Barriers in the form of limited resources remain a constant cause of hindrance to quality eyecare access among rural areas. However, with developments in AI and telehealth services, the situation can gradually be improved. Tele-optometry services powered with AI algorithms for image processing can significantly reduce the need of on-site specialists for diagnosis and detection of diseases.14
These AI software and tools, which are processed with huge data of respective images of both normal and abnormal conditions, can help even beginners to diagnose conditions, classify the stages, and provide appropriate referrals as necessary. Coupling the same with telehealth services can add human-based intelligence in framing the appropriate treatment plans in a collective, multi-disciplinary approach.
One of the most notable and significant examples of AI integration is the use of deep learning algorithms like IDx-DR in the screening of DR. Being FDA-approved, it has also proved efficient in screening DR and DME.15
When it comes to cases that need urgent care or treatment at tertiary centers, AI tools can triage conditions to vision-threatening, referrals, and prescription or counseling options.16 This community screening data, with proper consent and ethical considerations can also serve as a means for proper research and analysis leading to development of new health plans and initiatives in the community.
Caveats to using AI in primary eyecare
While AI can substantially improve the efficiency, ease and precision of not just eyecare, but overall healthcare, it comes with significant challenges and considerations. These caveats fall under various categories.
Patient confidentiality
The ethical concerns it comes with, is one of the major causes of its restricted adoption in spite of the other advantages. Data handling and privacy are severely compromised. The breach of privacy in sensitive patient information eventually hinders the acceptance of AI adoption in healthcare services.
Data bias
The success of AI integration depends on the data that is used to program these algorithms. For instance, an AI tool that is built with huge data of Caucasian population cannot be implemented in an area with predominant Hispanic population.
Accountability
Over-dependence on AI tools also brings out the issue of accountability and liability. When outcomes based on AI decisions are unsuccessful, improper accountability for the same, is a major concern. Without proper standards towards defining the responsibility in such cases, AI usage in healthcare services remains controversial.
Human vs. AI judgement
AI based decisions can sometimes differ from the standard experience-based human judgements. This can eventually lead to conflicting responsibilities and further compromise the care. Maintaining a proper balance between the two is crucial. AI should be treated as a supportive tool, not a replacement, especially in healthcare.
Investment challenges
Integrating AI, especially in primary eyecare centers, requires huge investments in terms of technology, infrastructure and training. These financial demands are one of the topmost challenges towards the practical implication of AI.
Societal implications
Automation of the administrative aspects, including that of scribing roles can lead to the reduction of job opportunities. This eventually leads to the risk of unemployment and other subsequent societal implications. The digital divide between those with access to AI and those who don’t could also leave an impact in the equity of the eyecare delivered.
Looking to the future
While AI has reached noteworthy advancements, the future still holds new promises towards its further development. More customized treatment plans, advanced predictive analysis, Virtual-reality (VR)-based visual enhancements for the visually impaired, and wearable technologies for rehabilitation are to name a very few in the near future.
With increasing awareness of AI usage, standard protocols and regulations are also being framed to address the ethical challenges of AI. Overcoming these challenges will essentially ensure the usage of AI to its maximum potential and enhance healthcare and eyecare. To utilize the benefits of AI integration to the fullest, it is crucial to stay updated on its advancements.
The latest trends and advancements in AI in eyecare are constantly updated in the new e-journal, AI in Eye Care.
Conclusion
There is no debate about the potential of AI tools. While AI can transform primary eyecare by increasing the precision, streamlining the process and addressing limited resources, it also comes with challenges that cannot be ignored.
Addressing these concerns and integrating AI tools with proper protocols are the future that eyecare professionals need to work towards.
