Boarding the Smart Future

While multiple artificial intelligence (AI)-based applications are already relevant for refractive surgeons, users must understand how these programs work and their limitations, advises David Smadja MD.

“We are being flooded by AI today, and there is no doubt that AI can help to improve efficiency, safety, and accuracy in refractive surgery,” Dr Smadja said. “We must be careful, however, not to over-rely on AI. Although the AI wave is here, we need to learn how to surf it.”

He proposed that busy refractive surgeons wanting to improve practice workflow and save time with surgical planning and patient care without compromising outcomes may find AI a valuable assistant in several ways. For example, a recently developed AI-enabled automated eye health platform collects and interprets subjective symptoms from the patient, conducts a comprehensive ophthalmic exam, and generates a report with guidance on choosing the best approach for individual patients.

Many teams have demonstrated good accuracy for AI approaches that predict subjective refraction based on a single aberrometry imaging acquisition. As another time-saving application, AI could be applied to develop questionnaires to triage a need for postoperative follow-up.

“We have all faced the situation where a patient calls a few days after PRK, wondering if blurry vision is normal,” Dr Smadja said. “What if the patient could use a smart tool at home that could answer the question automatically and either reassure the patient or raise a red flag that a visit is needed?”

AI-based programs are also helping refractive surgeons interpret data collected to screen patient appropriateness for refractive surgery. Discussing this application, Dr Smadja referred to the challenges of reviewing complex topography readouts where surgeons need to recognise patterns, detect abnormal values, and determine whether the latter are real or acquisition artefacts.

“We still have difficulty discriminating if a normal cornea is a keratoconus suspect or if what appears to be an early keratoconus is a normal eye,” he noted.

Yet while most corneal imaging tools provide AI-based scoring systems for guiding these decisions, Dr Smadja cautioned the need for understanding normality is relative because the determination depends on the algorithm’s training database, the features of which can vary across different patient populations.

“Normal ranges need to be adapted to the population seen in your clinical practice,” Dr Smadja said.

Other AI-based algorithms can guide patient selection for different refractive surgery procedures, integrating all clinical and demographic data for an individual or providing guidance for achieving the targeted refractive outcome. In addition, software created using AI can help to identify whether a patient is at risk for postoperative myopia regression, thereby allowing counselling about the possible need for a future enhancement.

Finally, refractive surgeons wanting to publish their research can find AI as a useful and time-saving aid for manuscript preparation.

“AI can help create a better manuscript—one standardised to the journal’s specifications by performing reference verification and checking for issues such as proper grammar, clarity, and formatting,” Dr Smadja said.

Dr Smadja spoke at the 2023 ESCRS Congress in Vienna.

David Smadja MD is the Director of Refractive Surgery, Hadassah Medical Center, Tel Aviv, Israel, and president of the Israeli Society of Refractive Surgery. smadj.david@gmail.com