AI for nAMD: Capturing more information from OCT


During a presentation at the EURETINA 2021 Virtual Congress, Dr Anat Loewenstein discussed how artificial intelligence can optimize optical coherence tomography.

AI for nAMD: Capturing more information from OCT

Reviewed by Dr Anat Loewenstein.

In a presentation at the EURETINA 2021 Virtual Congress, Dr Anat Loewenstein discussed optimizing optical coherence tomography (OCT) and how physicians can determine with even more accuracy precisely what is happening in patients’ eyes with neovascular age-related macular degeneration (nAMD) because of the potential afforded by application of artificial intelligence (AI).

Loewenstein is professor and director of the Department of Ophthalmology at Tel Aviv Medical Center, the Sidney A. Fox Chair in Ophthalmology, and vice dean, Sackler Faculty of Medicine, Tel Aviv University, Israel.

OCT is a major step forward in patient diagnosis, treatment, and monitoring but shortcomings remain. For example, physicians routinely make qualitative assessments of the presence and degrees of intraretinal/subretinal fluid and pigment epithelial detachments, but these are not precise assessment that are likely to result in poor inter-grader agreement and intra-grader consistency; OCT also provides the central subfield thickness, but the retinal fluid and neural tissue are not considered separately.

As Loewenstein pointed out, In neovascular AMD, it is important to distinguish between retinal fluid localization in the intraretinal and subretinal compartments and their volumetric information for informing retreatment decisions and predicting visual outcomes.

One technology to quantify OCT data, according to Loewenstein, is the Notal home-based OCT Analyzer (NOA), a machine learning algorithm that distinguishes normal morphologic features from elevated or distorted contours to quantify macular fluid of OCT volumes. The NOA quantifies intraretinal/subretinal fluid on Spectralis and Cirrus spectral-domain OCT devices. Both algorithms display their respective fluid output in nanoliters, which allows repeatable measurements and precise monitoring of disease activity.

“Recent advances in machine learning and deep learning have resulted in development of software that can process routinely acquired spectral-domain OCT data and precisely determine the location and volumetric information of macular fluid within different tissue compartments,” she said.

The software then can automatically generate multiple quantitative metrics related to macular fluid variables that may provide substantial advantages to research and clinical practice.

According to Loewenstein, using volume instead of thickness, separating fluid volume from neural tissue volume, distinguishing between intraretinal and subretinal fluid and monitoring dynamics over time might improve OCT’s ability to predict visual acuity changes.

Loewenstein emphasized the following:

• Because of the importance of OCT in assessing nAMD parameters, for which the fluid volumes in the retina are generally the key biomarkers, the availability of software that can analyze them quantitatively, automatically, and at scale is paramount.

• In practice, the ability to extract quantitative metrics of exudative activity from OCT scans is an important advantage over qualitative descriptions.

• By eliminating low intra-grader consistency and inter-grader agreement, exudative activity can be compared more meaningfully between sequential visits and visits that are widely separated of the same patient, or between patients, regardless of the individual physicians’ discretion.

• Quantitative metrics may result in better recordkeeping and visualization of exudative activity; the values from each visit can be recorded facilitating graphs of fluid volume plotted over time based on changes in treatment intervals, drugs, and doses.

• AI methods allow assessments and comparisons to be made separately for intraretinal/subretinal fluid and pigment epithelial detachments.

• Not all exudative activity is equal regarding visual prognosis and determining retreatment decisions.

• Future studies may lead to refined retreatment protocols possibly incorporating more nuanced approaches according to fluid compartment/volume.

• Algorithms that identify and quantify fluid types should help facilitate these approaches.

“AI tools may lead to more tailored treatment, ideally so that individuals achieve optimal visual outcomes with as few injections and visits as possible,” Loewenstein concluded. “This might help address some of the shortfall in visual outcomes that has consistently been observed between real-world practice and clinical trials.”


Anat Loewenstein, MD, MHA
Loewenstein is a consultant to Notal Vision.

See more EURETINA coverage

Related Videos
ARVO 2024: Andrew D. Pucker, OD, PhD on measuring meibomian gland morphology with increased accuracy
 Allen Ho, MD, presented a paper on the 12 month results of a mutation agnostic optogenetic programme for patients with severe vision loss from retinitis pigmentosa
Noel Brennan, MScOptom, PhD, a clinical research fellow at Johnson and Johnson
ARVO 2024: President-elect SriniVas Sadda, MD, speaks with David Hutton of Ophthalmology Times
Elias Kahan, MD, a clinical research fellow and incoming PGY1 resident at NYU
Neda Gioia, OD, sat down to discuss a poster from this year's ARVO meeting held in Seattle, Washington
Eric Donnenfeld, MD, a corneal, cataract and refractive surgeon at Ophthalmic Consultants of Connecticut, discusses his ARVO presentation with Ophthalmology Times
John D Sheppard, MD, MSc, FACs, speaks with David Hutton of Ophthalmology Times
Paul Kayne, PhD, on assessing melanocortin receptors in the ocular space
Osamah Saeedi, MD, MS, at ARVO 2024
© 2024 MJH Life Sciences

All rights reserved.