A retrospective cohort study has found that optical coherence tomography (OCT), which uses a deep learning algorithm, was able to effectively predict glaucoma risk.* The researchers looked at retinal nerve fiber layer thickness using fundus photographs—also known as visual photographs—to determine if subjects might develop visual field defects in the future.

What they did

Researchers analyzed 1,072 eyes of 827 glaucoma suspects with an average of about four fundus photographs per eye. They looked specifically at retinal nerve fiber layer (RNFL) thickness measurements over time. 

They reported that 196 eyes (18%) converted to glaucoma during follow-up (defined as converters). 

What they found

By using a machine-to-machine (M2M) OCT-trained deep learning algorithm, they were able to detect any glaucomatous damage before any visual field effects surface. 

Examples of this found by the algorithm include:

• The average baseline RNFL thickness was 88.7 microns (µm) for converters and 92.1 µm for non-converters.
• The mean rate of change for predicted RNFL thickness was faster for converters at -1.02 µm per year compared with -0.67 µm per year for non-converters, which is considered a statistically significant difference.
• There was a 71% increase in the hazard of developing visual field defects in RNFL for each 10 µm thinner in predicted baseline RNFL thickness.
• Faster decreases of 1 µm per year predicted RNFL thickness was associated with about a twofold increase in developing visual field loss.

From these findings, researchers deduced that “eyes were significantly more likely to convert to glaucoma if they had lower M2M predictions of RNFL thickness at baseline and faster declines in thickness over time.”

What this means

The study’s co-author,  Felipe A. Medeiros, M.D., Ph.D., explains that “these [RNFL thickness]  measurements could help clinicians decide which suspects are at higher risk for developing glaucoma and may require treatment or closer monitoring,”

This new algorithm is effective for population-based screening for glaucoma and provides an opportunity for screening in a primary care setting. Since the algorithm doesn’t require access to OCT instruments, it is versatile and can be considered a cost-effective and accessible test.

Since the method focuses primarily on simple fundus photographs, however, it is not meant to replace OCT. “[I]n the absence of OCT or when its use is impractical, the method could achieve great results,” said Medeiros.

Future research must look at the validity of the algorithm’s use in other populations.

*Linnehan, R. (2021, Mar. 16). Deep learning algorithm’s prediction of RNFL thickness gauges risk for glaucoma conversion. Ocular Surgery News. https://www.healio.com/news/ophthalmology/20210316/deep-learning-algorithms-prediction-of-rnfl-thickness-gauges-risk-for-glaucoma-conversion