OCT-based AI analysis brings changes in GA and therapeutic responses into sharp focus

(Image credit: AdobeStock/Shuo)

Researchers have been delving deeply into geographic atrophy (GA) to increase the knowledge base of end-stage age-related macular degeneration (AMD). The focus has been on earlier intervention to identify and prevent progression of GA.

The introduction of artificial intelligence (AI) into the investigative process is giving researchers an edge by both simplifying their efforts in imaging analysis and providing more precise retinal details in patients with GA secondary to AMD.

Identification of disease activity and therapeutic response in GA in the FILLY study

Ursula Schmidt-Erfurth, MD, reported the results of a post hoc analysis of an optical coherence tomography (OCT)–based AI study at the 2024 American Society of Retina Specialists Annual Meeting in Stockholm, Sweden, that prospectively analyzed the natural progression of GA in the FILLY randomized clinical trial (NCT02503332) in participants treated with pegcetacoplan therapy (Syfovre; Apellis Pharmaceuticals). She is a professor and the chair of the Department of Ophthalmology at the Medical University of Vienna in Austria.

In that study, ophthalmologists sought to predict the yearly growth rate of GA and to select the potentially faster-growing lesions from 2 eyes based on fundus autofluorescence (FAF), near-infrared reflectance (NIR), and OCT. A deep learning algorithm predicted progression solely on the baseline OCT (Spectralis; Heidelberg Engineering).

A total of 134 eyes of 134 patients from the phase 2 clinical trial were included, and 2880 gradings were performed by 4 ophthalmologists. The main outcomes were the accuracy, weighted κ, and concordance index (c-index).

Schmidt-Erfurth reported, “Human experts reached accuracy values of 0.37, 0.43, and 0.41, and κ values of 0.06, 0.16, and 0.18 on FAF, NIR+OCT and FAF+NIR+OCT, respectively. A pairwise comparison task showed that human experts achieved c-indices of 0.62, 0.59, and 0.60.”

In comparison, the automated AI-based analysis reached an accuracy of 0.48, a κ value of 0.23 on the first task, and a c-index of 0.69 on the second task using only OCT imaging.

She pointed out that although the human gradings have improved with the availability of OCT, AI performed in an automated manner that is proving to be superior to and faster than analyses performed by ophthalmologists.

The advantages of AI are that
the technology is fully automated and provides reliable analysis of routine OCT as well as being accessible at any place and any time through the cloud.

AI performance in a post hoc analysis of OCT in the OAKS and DERBY trials of pegcetacoplan for GA

Another recently published report1 also underscored the contribution of AI technology to increasing the knowledge about GA.

Schmidt-Erfurth was the lead author of a post hoc OCT-based AI analysis performed to identify changes in the mean area of retinal pigment epithelial (RPE) loss and ellipsoidal zone (EZ) loss over time. The study results showed that OCT-based AI analysis objectively identified and quantified degeneration of the photoreceptor and RPE in patients with GA secondary to AMD.

The investigators wanted to quantify the morphologic changes of the photoreceptors and RPE layers in patients with GA treated with pegcetacoplan therapy. Patients with GA due to AMD had been participants in 2 prospective randomized phase 3 clinical trials (OAKS, NCT03525613; DERBY, NCT03525600). Spectral-domain OCT images were analyzed over 24 months for changes in the mean area of RPE loss and EZ loss over time in the pooled sham arms and the monthly (PM)/every other month (PEOM) treatment arms.

Analysis of results

A total of 897 eyes of 897 patients were included. At 24 months, the analysis showed a therapeutic reduction of RPE loss growth by 22% and 20% in the OAKS trial and 27% and 21% in DERBY for PM/PEOM compared with the sham arm, respectively.

The authors also reported that the reduction on the EZ level was significantly higher with 53% and 46% in the OAKS trial and 47% and 46% in the DERBY trial for PM/PEOM compared with sham at 24 months.

They commented, “The baseline EZ-RPE difference had an impact on the disease activity and therapeutic response. The therapeutic benefit for RPE loss growth increased with larger EZ-RPE difference quartiles from 21.9%, 23.1%, 23.9% to 33.6% for PM vs sham (P < 0.01 for all comparisons ) and from 13.6% (P = 0.11), 23.8%, 23.8% to 20.0% for PEOM vs sham (P < 0.01 for all comparisons) in quartiles 1, 2, 3, and 4, respectively, at 24 months. Regarding maintenance of the EZ layer, the therapeutic reduction of loss increased from 14.8% (P = 0.09), 33.3%, 46.6% to 77.8% (P < 0.0001) between PM and sham and from 15.9% (P = 0.08), 33.8%, 52.0% to 64.9% (P < 0.0001) between PEOM and sham for quartiles 1 to 4 at 24 months.”

Based on their findings, the authors concluded that “OCT-based AI analysis objectively identifies and quantifies PR and RPE degeneration in GA. Reductions in further PR degeneration consistent with EZ loss on OCT are even higher than the effect on RPE loss in the phase 3 trials of pegcetacoplan treatment. The EZ-RPE difference has a strong impact on disease progression and therapeutic response. Identification of patients with higher EZ-RPE loss difference may become an important criterion for the management of GA secondary to AMD.”

Schmidt-Erfurth and colleagues also pointed out that AI-based clinical tools will become widely available via cloud-based technology. “… [This] further facilitates real-time and ubiquitous access to a therapeutic option for one of the most severe diseases in our societies. Our research effort undertaken in this context may add another step for envisioning clinical end points and medical devices that eventually will benefit providers, health care systems, and patients.” •

Reference

Schmidt-Erfurth U, Mai J, Reiter GS, et al. Disease activity and therapeutic response to pegcetacoplan for geographic atrophy identified by deep learning-based analysis of OCT. Ophthalmology. Published online August 14, 2024.doi.org/10.1016/j.ophtha.2024.08.017