People often view the current generation of vision correction as the fourth. However, we proudly proclaim that we are in the fifth generation of vision correction, considering the fact that we have gone beyond the distinction of generations with regard to both experience and equipment and have reached an age when artificial intelligence plays a major role in vision correction surgery.
Vision correction is divided into separate generations based on two criteria. The first criterion is based on the surgical method; typically, laser epithelial keratomileusis (LASEK), laser-assisted in situ keratomileusis (LASIK), phakic intraocular lens implantation (ICL), and small incision lenticule extraction (SMILE) as the first, second, third, and fourth generation, respectively. Vision correction surgery has evolved from one generation to the next with methods designed to reduce the extent of incision and pain. Classifying generations by surgical methods is simply a comparison of equipment since the key determining factor is which equipment will be used to make how much of an incision on the cornea. Currently, even multifocal IOL for cataract is within the scope of vision correction. Since it is possible to correct aging-related vision regression, it is certainly not an overstatement to claim that technology has advanced as much as in could. Another criterion categorizing vision correction into generations is stability: experience, equipment, Avellino genes, and big data integrating all genetic and clinical information as the first, second, third, and fourth generation, respectively. Vision correction has advanced beyond merely more experienced physicians and more precise and safer equipment, and has reached the point of examining genes. It has advanced in the direction of screening for eyes that should not undergo vision correction surgery in correcting corneal abnormalities when maximum prevention of corneal damage is imperative, such as keratoconus and Avellino gene. Considering that the advances in surgical methods and equipment were focused on increasing the safety of vision correction surgery, the key element for vision correction surgery should be safety.
AI will raise the scientific standard to the highest level with regard to surgical safety. Concurrently, its introduction has triggered several scholars, globally, into publishing their study findings on the importance of screening for eyes that need to avoid undergoing vision correction surgery with AI-assisted. This is because AI, based on vast amount of data, can present the most suitable surgical method and cost for the patient’s eyes regardless of experience, availability of equipment, and financial preference.
A learning curve refers to a pattern of sluggish growth rate relative to the amount learned in the beginning, followed by advances in leaps and bounds at some point onward. This is often experienced when learning a foreign language. This also represents great deal of growth at some point based on steadily accumulated experience and mistakes. Even surgeons face a learning curve. Surgeons with only 2-3 years of surgical experience can produce good surgical outcomes. However, they may be considered relatively inadequate when compared to surgeons with 7-8 years of experience. Meanwhile, surgeons with experience that spans a period of 10-15 years or more could perform surgeries with an instinct, in addition to the associated test results. Fundamentally, nothing can overcome steady accumulation of experience.
If that is the case, do novices always have to trail those who are experienced? Is there no way to quickly acquire the experience and mistakes of someone else? Today, we are able to absorb much diverse knowledge and information at a faster pace than those from a century ago, owing to the experience and knowledge of our predecessors, who left the same behind in the form of books. Today, anyone can access such experience and knowledge through the Internet.
Would this be possible for medicine as well? Even after graduating from medical school and becoming a specialist, ophthalmologists must learn from the very beginning to be able to actually claim their place in the field of surgery. Furthermore, it requires a tremendous amount of time and effort. However, there are places, which lack proper medical facilities and equipment, making it difficult to even find the opportunity to invest time and effort to accumulate experience. At B&VIIT, we actively conduct medical volunteering programs each year. During such programs, we perform a large number of surgeries within a short period. The places we go to for the medical volunteer work do not usually have high prevalence of eye diseases, which could primarily be attributed to medically vulnerable regions lacking advanced medical facility and equipment owing to economic reasons; therefore, the patients’ conditions worsen and the physicians lack the opportunity to accumulate adequate experience. Consequently, many cases occur in which surgery cannot be performed despite prior knowledge of the disease.
Therefore, if our experience form our dedication to vision correction, even if it were to be as short as a few years or as long as several decades, could be transferred to other physicians, it would be possible to reduce the cost of fostering new physicians and regions isolated from medical benefits. Therefore, we began our efforts to develop an AI, in the hopes that it would help physicians who lack experience to perform error-free surgeries similar to the veteran physicians. It has been a few years since we began to invest in AI, and much to our immense satisfaction, AI from B&VIIT is capable of identifying the knowledge and subject matter expertise of each physician, making us wonder “how did it know about this?” Although it was amazing at first, but as we move forward, we realized that we wanted to do much more with AI.
The desire to do more with AI was based on its capability of data analysis at a rate that is humanly not possible to achieve. Ophthalmologists are often limited by the number of test results and images they are required to review within a short time. Moreover, there are parameters that cannot be analyzed by the human eyes; for example, those that physicians cannot differentiate, regardless of their experience, could be differentiated through deep learning, including factors like accurately identifying the gender, blood pressure, and smoking habits of a patient simply by looking at images of the retina.
We can get a glimpse of the future through its potential in identifying abnormalities that humans cannot and alerting us to situations that would not otherwise cross our minds. AI could also play a role in resolving issues that cannot be overcome with currently available medicine – curing fatal diseases or resolving or improving dry eyes and glare that people were forced to adapt to despite the associated discomfort.
AI from B&VIIT is capable of selecting the optimal surgical method; accurately calculating the surgical cost and size of the IOL(intraocular lens); and precisely predict the expected postoperative visual acuity; such is its level now. Although it produces questionable results occasionally, it could still be used by experienced physicians as a system to revalidate their results and as a reliable reference for physicians who have just started performing surgeries.