The patients and doctors have different questions when they first meet at the doctor’s office. The thing that the patient is most curious about is ‘How well will I see after the surgery?’. In contrast, the first question that doctor has is ‘Is this the eye condition that needs surgery?’ or whether surgery is possible. What B&VIIT AI discovers is also whether surgery is possible. The most important thing is to screen the eye conditions that can not be permissible for the surgery.
There are eyes that should never receive corneal surgery. Keratoconus is the representative case. A normal cornea has smooth curve like the oval shaped American football ball, but keratoconus has bulging type of cornea like a cone. Therefore, vision becomes blurry such as serious myopia and astigmatism, and the night vision becomes dark and sensitive to light. Candidates with keratoconus can correct it through glasses or contact lens, but in the worst case, they may lose their vision or have to receive corneal transplant.
There are cases where the form of cornea is changed like keratoconus after vision correction surgery. This is called corneal ectasia that makes the inner layers of your cornea become weak, causing the cornea to change shape, protrude forward and distort your vision. Naturally, severe myopia and astigmatism, such as keratoconus, occurs and if it were more severe, it may require corneal transplant or may lose vision.
The corneal ectasia was a side effect that occurs many times in the beginning of LASIK surgery. As long as LASIK surgery cuts cornea and corrects vision, the potential danger of corneal ectasia can always exist. This is the irony of the possibility of losing the existing vision after receiving vision correction surgery to see better.
In this way, it is not excessive to say that the first thing the ophthalmologist must consider is avoiding eyes that should not receive surgery. Thus, the first question of doctor and the first response of B&VIIT is whether surgery is possible or not possible.
As the standard of selecting surgery method and deciding the possibility of surgery is stability, the research direction of AI is also focusing on stability. This is not only the standard of B&VIIT Eye Center, but also the overall research propensity of ophthalmology AI. Globally, research results of asserting the use of AI to distinguish eyes that cannot receive vision correction surgery are being published continually.
Professor Renato Ambrosio Jr. of Brazil’s Rio University, through the contributions published in <Seminars in Ophthalmology> on January 2019, discussed the corneal ectasia after LASIK surgery and asserted that machine learning and algorithm can ply a useful role in checking keratoconus. The ‘Tomographic Biomechanical Index’ (TBI), which has taught the corneal 3D tomography and biomechanical information through machine learning, has its results loaded on the testing machine and can find if the cornea is not suitable for vision correction surgery through the corneal test.
<Ophthalmology>, the academic journal of AAL(American Academy of Ophthalmology), is an authoritative magazine which includes ophthalmological knowledge with a new theory. It states that AI must be used to distinguish keratoconus where LASIK surgery cannot be conducted. Therefore, it is proposing that there is need for more active research using AI.
This is not the first time that it has emphasized the need for AI in assessing whether surgery is required. There was already a research thesis in 2016 that asserted that corneal topography test result was taught through machine learning and discover keratoconus. To add, TBI, the index using machine learning explained above was verified to have very high clinical precision.
What many researches commonly say is that people who cannot receive surgery or people who have high possibility of having poor outcomes associated with the surgery should be identified through machine learning. It is no exaggeration to say that the core of research regarding AI in vision correction is stability.
There are two reasons for this. The first is eyes with keratoconus should not receive surgery and corneal ectasia is the most lethal complication of vision correction. In both cases, conducting surgery to improve vision has the danger of complete loss of vision. The second is that distinguishing the test result can be better performed by AI. Conducting diverse tests allows evaluating many indices that the ophthalmologist must use. The test results involve analyzing complex results such as not only analyzing if surgery is possible, but if optimal vision after surgery is achievable and the possibility of complication after surgery.
The parts that must be examined and considered by the doctor can be seen in one go by AI. It is the same principle as computer doing complex calculations better. Moreover, due to the characteristics of ophthalmologists, where there are many cases of test results being expressed in the form of an image, the AI learnt through image deep learning can also show in one go the test result that needs a long time for ophthalmolgists to assess.
In fact, the first AI of B&VIIT distinguished the danger of keratoconus or corneal ectasia with high precision.
Identifying eyes that cannot receive surgery, by the means of machine learning that learnt data of millions of people, was the same as distinguishing with experience of doctors conducting surgery using data that had been previously amassed data. It is because not making a mistake that misses the case of eye with potential complications is more important than having excellent surgical proficiency.