We are pleased to share an overview of our published abstract in the IOVS Journal (June 2024). This study shows the correlation between Geographic Atrophy (GA) lesions identified by fundus autofluorescence (FAF) and morphological changes in the outer retina using Optical Coherence Tomography (OCT). Dive into the detailed analysis below and discover how our innovative AI-based OCT segmentation provides deeper insights into retinal layer changes in the context of  GA assessment.

‍

Introduction

Currently, the area of Geographic Atrophy (GA) lesions and the growth rate assessed by fundus autofluorescence (FAF) are the gold standard for the morphological assessment of GA in clinical practice and clinical trials ¹. However, FAF shows some important limitations ² and one of them is the restricted information about the depth of the lesion and morphological changes at a retinal-layer level. In this scenario, Optical Coherence Tomography (OCT) plays a pivotal role since it allows a three-dimensional study of the atrophic lesion, providing detailed and quantitative information on the loss of specific retinal layers.

‍

Methods

We investigated the correlation between hypofluorescent lesion areas in FAF and the extent of retinal pigment epithelium (RPE), Ellipsoid Zone (EZ), Outer Photoreceptor Segment (OPR), and Interdigitation Zone (IZ), and myoid zone (MZ) loss on OCT, along with the presence of abnormal hyperreflective areas (amorphous material, AM). In collaboration with Prof. Dr. Dr. Marion Munk and Dr. Irmela Mantel, we retrospectively analysed OCT cubes (N=180) of patients (N=79) with GA secondary to Age-related Macular Degeneration (AMD) obtained at the Medical Retina Department at the Jules-Gonin Eye Hospital (Lausanne, Switzerland). 

‍

Hypofluorescent lesions on FAF images were manually labelled by expert graders and automatically aligned to infrared images to match GA areas in OCT for Dice score evaluations. We used our artificial intelligence-based algorithm (RetinAI's OCT Segmentation model, model for Research Use Only) to automatically segment and analyse the retinal layer thickness at a B-scan level (Figure 1-A). This module is integrated with RetinAI Discovery® platform. 

We then calculated en-face areas of structural loss of MZ, EZ+OPR+IZ, and RPE (Figure 1-B), which were then correlated with FAF-measured GA areas using Person's coefficient. 

‍

Figure 1. A- OCT segmentation model featuring the segmentation of seven layers of the retina, the choroid, fluids and other relevant biomarkers. B- Defect maps for photoreceptor layers and RPE. 

‍

Results

As shown in Figure 2-A, the hypofluorescent lesion area on FAF correlated with areas of RPE loss (correlation coefficient: 0.96). The mean and standard deviation measured between FAF (hypofluorescent area) and OCT (RPE loss) resulted in 0.258 ±1.28 mm² (Figure 2-B). Mean Dice score was calculated between areas of RPE loss and registered FAF with values of 0.731 (0.193).  In addition, we found a correlation between the atrophic lesion in FAF and EZ+OPR+IZ and MZ loss (Correlation coefficients: 0.94 and 0.93, respectively). This is consistent with the retinal pigment epithelium (RPE) and outer retinal atrophy (RORA) definition in OCT proposed by the CAM group ³. Finally, we observed moderate positive correlations between the FAF GA area and AM area (0.65). 

‍

Figure 2. Correlation between FAF and OCT measurements. A- FAF lesion area positively correlated with RPE, PR, and MZ loss in OCT. B- Comparison of FAF hypofluerescent area with RPE loss EZ+OPR+IZ loss and MZ loss areas. 

‍

Interestingly, the areas of EZ+OPR+IZ loss and MZ loss were larger than the hypofluorescent lesion in FAF and the RPE loss (EZ+OPR+IZ loss area: 6.04 mm² for, MZ loss area: 5.86 mm², RPE loss area: 4.45 mm², FAF: 4.70 mm²) (Figure 2, B). These results suggest that there are significant changes in the photoreceptor layers beyond the actual GA lesion as measured on FAF, which may explain the topographic evolution of the lesion. We further explored this hypothesis as part of another project, which can be found in our detailed analysis of retinal layer thickness surrounding GA lesions. 

‍

Conclusions

Overall, our results suggest that OCT and FAF measurements show a good correlation and that the assessment of RPE and PR loss in OCT represent reliable biomarkers for GA assessment that could be employed in pre-screening and screening in clinical trials and, later on, in clinical practice. 

With our new RUO OCT Segmentation Model, elevating retinal disease analysis is possible! Our model is designed to make your research and clinical workflows more efficient and insightful. With our platform RetinAI Discovery®, you will have access to powerful tools to visualize, quantify and predict disease progression with precision.

Want to explore how our platform can transform your R&D initiatives and enhance patient outcomes? Request a demo, and see our technology in action! We are here to enable the right decisions sooner in Healthcare.

‍

‍

References

1. Sadda SR, Chakravarthy U, Birch DG, et al. Clinical Endpoints for the Study of Geographic Atrophy Secondary to Age-Related Macular Degeneration. Retina (Philadelphia, Pa). 2016;36:1806. doi:10.1097/IAE.0000000000001283

2. Birtel J, Bauer T, Pauleikhoff L, et al. Fundus autofluorescence imaging using red excitation light. Sci Rep. 2023;13:9916. doi:10.1038/s41598-023-36217-x

3. Sadda SR, Guymer R, Holz FG, et al. Consensus Definition for Atrophy Associated with Age-Related Macular Degeneration on OCT: Classification of Atrophy Report 3. Ophthalmology. 2018;125:537-548. doi:10.1016/j.ophtha.2017.09.028

‍