All eyes on the future: How AI is revolutionizing ocular surgery

Image Credit: AdobeStock/ipopba

Artificial intelligence (AI) has been evolving so quickly that by the time this article has been published, there will already be significant new developments and research to be explored in eye care. Among the most discussed AI developments recently is ChatGPT, created by OpenAI. Introduced to the public in 2022, it is a type of AI developed for continuous learning and designed to converse with its users. All AI systems continually refine their algorithms, offering limitless possibilities in their applications. Currently, AI is used in eye care for evaluating retinal images for detection of diabetic retinopathy (DR). This parallels the advancements expected in refractive and cataract surgery. This article will explore the current utilization of AI in ocular care and surgery and speculate on future developments.

Defining AI

To have a more informed conversation regarding AI, it is important to understand what it is and what it is not. According to IBM, the definition of AI is the theory and development of computer systems able to perform tasks and cognitive functions that normally require human intelligence, enabling them to learn, read, write, create, and analyze.¹ There are 7 types of AI: narrow AI, artificial general intelligence, artificial superintelligence, reactive machine AI, limited memory AI, theory of mind AI, and self-aware AI.²

Narrow AI, or weak AI, involves tools meant to perform specific tasks; examples include self-driving cars and AI virtual assistants.2 Artificial general intelligence is designed to perform multifunctional tasks and provide humans with equally intelligent assistants, such as ChatGPT (a form of generative AI) and quantum hardware.² As for the AI utilized by Netflix for recommendations, that is reactive machine AI. The next milestone in AI would be theory of mind, which is the concept of AI that can perceive human emotions and predict actions based on its findings.²

And the autonomous type we associate with futuristic movies? That’s self-aware AI, which not only can sense feelings of others but also will have a sense of self and its own feelings.² The AI currently employed for DR detection utilizes limited memory AI and deep learning techniques, mirroring the functionality of neurons in the human brain. This enables the technology to assimilate information and data from various experiences, facilitating continuous learning and enhancing accuracy progressively over time.²

Current uses of AI

AI is being used to help optimize outcomes and minimize complications in refractive and cataract surgeries. In refractive surgery, AI aids in screening candidates and predicting postoperative outcomes. Machine learning (ML) models, such as those developed by Yoo et al, automate the screening process for potential candidates, demonstrating performance similar to that of eyecare professionals.^3,4 Xie et al introduced the Pentacam InceptionResNetV2 Screening System for identifying suitable candidates, achieving an impressive detection accuracy of 94.7%.³ Additionally, AI models developed by Yoo et al and Kang et al select expert-level surgical options and predict postoperative implantable collamer lens (ICL) vault, respectively, demonstrating promising results in improving surgical decision-making and outcomes.^3,4

ML techniques, such as random forest regression, predict postoperative ICL vault accurately, reducing the risk of complications.³ Furthermore, deep learning methods, like those developed by Sun et al, enable the accurate monitoring of ICL position postoperatively, enhancing clinical outcomes.³ In predicting postoperative complications, AI models developed by Lopes et al and Kim et al effectively differentiate between susceptible cases and predict the risk of myopic regression after surgery, aiding in early intervention and improving patient prognosis.³

In cataract surgery, AI contributes to enhancing intraocular lens (IOL) power calculation accuracy and optimizing surgical techniques. AI-based IOL calculation formulas, such as the Clarke and FullMonte formulas, demonstrate enhanced accuracy compared with conventional formulas, improving patient outcomes.³ Additionally, robotic-assisted surgery systems, like the Interventional Artificial Eye Surgery System and real-time intraoperative guidance platforms, provide surgeons with precise instrument positioning and feedback to improve surgical safety
and efficacy.³

Future uses of AI

There is a significant unmet need when it comes to screening and treatment of corneal diseases and cataracts, especially in developing countries. AI algorithms, both imaging and nonimaging based, could facilitate timely diagnosis and treatment and advance areas in refractive surgery.

However, the utilization of AI for anterior segment diseases remains in its infancy, with several limitations to be addressed before implementation in clinical settings. Standardization of imaging techniques poses challenges due to variability in magnification and contrast, among other factors.⁴ Preparing high-quality data annotated by experts is time-consuming, and algorithms requiring vast data sets, such as convolutional neural networks, may need to be replaced with more self-supervised or unsupervised methods.4 Validation of large data sets is necessary with consideration for medicolegal and data security concerns.⁴

To ensure AI models identify relevant factors effectively, high-resolution images and accurate data input are crucial, alongside technological advancements like handheld retinal cameras and cloud computing.⁴ The “black box” effect, defined as the inability to identify how a deep learning system such as AI is making its decisions, requires further research. Additionally, standardized nomenclature has the potential to enhance reproducibility and generalizability.⁴

Further, clinical implementation and integration with big data and telemedicine hold promise for widening access to health care services.4 However, barriers such as cost, expertise, and regulatory issues must be addressed. Despite the high investment required for training and developing AI algorithms, it may be necessary to prevent strain on the global health care system we have seen post 2020 due to the COVID-19 pandemic. Multitasking models could improve cost-to-benefit ratios, improving efficacy and diagnosis and decreasing chair time.

Another application investigated by Kundu et al is the possibility of utilizing AI for identifying factors that influence patient decisions regarding refractive surgery.⁵ By employing AI-driven approaches, ophthalmologists could analyze extensive patient data, including demographics and medical history, to discern patterns that affect decision-making processes.5 This enables personalized care delivery tailored to individual preferences and concerns.⁵

Perception of AI involvement

This begs the question: Will the inclusion of AI in practice be well received? This loaded question does not have a single answer. With 7 different types of AI, the implication and application could be varied. A study surveyed 400 optometrists regarding their perspectives toward AI technology.6 It found that optometrists recognize AI’s capacity to streamline clinical workflows, improve diagnostic accuracy, and enhance patient care delivery.⁵ However, concerns regarding AI’s reliability, data security, and potential impact on professional autonomy are also highlighted.6 Ultimately, the study explores optometrists’ preferences for AI-assisted tools and investigates their readiness to embrace technological advancements in their practice. The study also emphasizes the importance of training and education to ensure optometrists’ competence in utilizing AI tools effectively.⁶

In conclusion, the evolving landscape of AI offers promising opportunities and significant challenges in ocular care and surgery. Current AI applications in optimizing surgical outcomes demonstrate effectiveness in enhancing patient care. Looking ahead, AI holds potential for addressing unmet needs in timely diagnosis and treatment of corneal diseases and cataracts, particularly in developing countries and underserved regions. However, challenges such as cost, standardization of imaging techniques, and data quality need to be addressed for successful implementation. Concerns about reliability, data security, and professional autonomy persist, emphasizing the importance of training and education. Embracing AI’s potential while addressing challenges is crucial for its responsible, effective integration into clinical practice. By doing so, we can harness AI’s transformative power to improve patient care and eye care delivery globally.

References:

1. What is artificial intelligence (AI)? IBM. Accessed May 22, 2024. https://www.ibm.com/topics/artificial-intelligence

2. Betz S, Whitfield B. 7 types of artificial intelligence. Built In. February 6, 2024. Accessed May 22, 2024. https://builtin.com/artificial-intelligence/types-of-artificial-intelligence

3. Nuliqiman M, Xu M, Sun Y, et al. Artificial intelligence in ophthalmic surgery: current applications and expectations. Clin Ophthalmol. 2023;17:3499-3511. doi:10.2147/OPTH.S438127

4. Rampat R, Deshmukh R, Chen X, et al. Artificial intelligence in cornea, refractive surgery, and cataract: basic principles, clinical applications, and future directions. Asia Pac J Ophthalmol (Phila). 2021;10(3):268-281. doi:10.1097/APO.0000000000000394

5. Kundu G, Virani I, Shetty R, Khamar P, Nuijts RMMA. Role of artificial intelligence in determining factors impacting patients’ refractive surgery decisions. Indian J Ophthalmol. 2023;71(3):810-817. doi:10.4103/IJO.IJO_2718_22

6. Scanzera AC, Shorter E, Kinnaird C, et al. Optometrist’s perspectives of artificial intelligence in eye care. J Optom. 2022;15(suppl 1):
   S91-S97. doi:10.1016/j.optom.2022.06.006