• COVID-19
  • Biosimilars
  • Cataract Therapeutics
  • DME
  • Gene Therapy
  • Workplace
  • Ptosis
  • Optic Relief
  • Imaging
  • Geographic Atrophy
  • AMD
  • Presbyopia
  • Ocular Surface Disease
  • Practice Management
  • Pediatrics
  • Surgery
  • Therapeutics
  • Optometry
  • Retina
  • Cataract
  • Pharmacy
  • IOL
  • Dry Eye
  • Understanding Antibiotic Resistance
  • Refractive
  • Cornea
  • Glaucoma
  • OCT
  • Ocular Allergy
  • Clinical Diagnosis
  • Technology

ARVO LIVE: Predictive AI algorithms and data interpretation


Ophthalmology Times® talked with T. Y. Alvin Liu, MD, about predictive AI and its uses in ophthalmology and screening of the eyes at this year's ARVO meeting.

Ophthalmology Times® talked with T. Y. Alvin Liu, MD, about predictive AI and its uses in ophthalmology and screening of the eyes at this year's ARVO meeting.

Video transcript

Editor’s note: Transcript lightly edited for clarity.

T. Y. Alvin Liu, MD:

This year at ARVO 2023. I was involved in an educational course on artificial intelligence. Specifically, I gave a talk on predictive artificial intelligence.

Most of the published studies to date in the field of ophthalmology and artificial intelligence revolves classification. What I mean by that is classifying whether a certain image contains referable diabetic retinopathy or not, or whether certain image contains glucometers changes in optic disk.

The new class of artificial intelligence models involve predictive tasks. and in my talk, I highlighted several categories of predictive tasks. The first one is Structure. Structure. Prediction. For example, predicting OCT metrics in eyes with diabetic macular edema using only color for this photograph.

The second category of predictive task is Structure. Function. Prediction. An example will be predicting changes in visual fields. Eyes with glaucoma, using color fundus photograph.

The third category of predictive tasks as a prediction of disease progression or onset. An example we'll be using OCT images to predict imminent conversion from dry to wet AMD within six months.

The last category of predictive tasks is predated predicting treatment response. and in this context, most of the published studies today have focused on predicting treatment response to anti-VEGF injections in conditions such as diabetic macular edema, or wet AMD.

So these are the categories of the newest studies that involve predictive tasks in artificial intelligence. There are many different implications for these studies. But I think the two main areas that are most exciting for these predictive model.

The first is these will see application and clinical trials, specifically in foot conditions where visual acuity may not be the best functional endpoint, I can see a day when objective reproducible images such as OCT, can replace some of these functional endpoints if we can consistently demonstrate that functional endpoints can be predicted from images alone. and the second application that I can see, that's very excited. That's very exciting, is using these predictive models to tailor or predict treatment response in patients with VEGF driven diseases.

For example, in the context of neovascular AMD, we could use these models predicts whether someone's going to need a lot of injections over the long run or not a lot of injections or we can use these models predicts the short term anatomic response in some of these patients. For example, the change in central subfield thickness in OCT from treatment initiation to six months onwards

Related Videos
EyeCon 2024: Peter J. McDonnell, MD, marvels on mentoring, modern technology, and ophthalmology’s future
© 2024 MJH Life Sciences

All rights reserved.