Mark Packer, MD sat down with Sheryl Stevenson, Group Editorial Director of The Ophthalmology Times®, to discuss his presentation on Machine Learning and Predicting Vision Outcomes after Cataract Surgery at the ASCRS Annual Meeting in San Diego. rice field.
video transcript
Editor’s Note: This transcript has been edited for clarity.
Cheryl Stevenson:
Dr. Mark Packer, who will be presenting at this year’s ASCRS, is also in attendance. Hello Dr. Packard. It was great to see you again.
Mark Packer, MD:
Nice to meet you, Cheryl.
Cheryl Stevenson:
Yes, can you tell us a little bit about your story about machine learning and visually predicting vision outcomes after cataract surgery?
Mark Packer, MD:
Indeed, as we know, humans are prone to mistakes, and surgeons don’t like to admit it, but they tend to make mistakes from time to time. One is to always extrapolate from the latest experience. So if you have one patient who is very dissatisfied with their multifocal intraocular lenses, suddenly you will be more cautious with the next patient, and probably the next patient after that.
Also, the opposite can happen. If I had a patient who was so excited about his toric multifocal that he would never have to wear glasses again and leave for Hawaii in the morning, completely transformed, I would think. It was the best thing I have ever done. And now suddenly everyone looks like a candidate. I have. It is just human nature.
And what we try to do with our ophthalmology program is bring a little bit of objectivity into the mix. We let the algorithms do the work. For example, instead of just looking at the refractive power of the lens, we actually look at the actual optical properties of the lens, the modulation transfer function, to help correlate with what the patient wants in terms of eyeglasses. independence.
But the real brainchild here is the idea of incorporating post-surgery patient feedback into the decision-making process. The VFQ-25 is a survey of visual function by the RAND company in the 1990s that measures how satisfied people are with their vision. We surveyed whether you should worry about it, how you feel about your vision, whether you are comfortable driving at night, and more. .
If I could incorporate that feedback into my decision-making, instead of going to the next room, I could just keep in mind what happened today and actually incorporate the knowledge of every patient I know. Ever since I started using this system, I have continued to perform surgeries.
The machine learning algorithm actually takes feedback from this patient and uses it to identify preoperative characteristics, such as personal items such as hobbies, what they do for recreation, what their job is, what they are like, etc. Can be combined with e.g. if you have visual requirements. Also, anatomical factors, axial length, anterior chamber depth, corneal curvature, etc., can all be brought together to really start tailoring interocular lens selection to the patient. We also investigated the patient’s personal characteristics and how they actually felt about the outcome of the surgery.
This is how I think machine learning can help us, and hopefully help surgeons understand premium IOLs more quickly. For premium lenses, especially multifocal extended depth-of-field lenses, there are visual side effects and limitations, but there are also significant advantages. So we hope that using machine learning will help young surgeons build confidence more quickly and increase patient adoption of these premium lenses.
