Abstract
Purpose :
Hyperspectral AF image analysis has been successful in separating key spectra ex vivo of human RPE FM in eyes with and without AMD.1 Comparison with CS AF imaging can further validate this data.
Methods :
Retinal cross-sections from two eyes from a single AMD donor underwent hyperspectral AF imaging at 40X in 10 locations total with 3 excitation wavelengths (436, 480 and 505 nm), and emissions recorded from 420 to 720 nm. The resulting image cubes were simultaneously decomposed with non-negative matrix factorization (NMF) to recover emission spectra and their corresponding localizations. These spectra were compared to spectra from en face images of pure RPE/Bruch’s Membrane (BrM) flatmounts from 19 subjects2.
Results :
The short wavelength spectral sources on cross section include the RPE, drusen, subRPE deposits and BrM, plus choroidal collagen and photoreceptors not present on RPE/BrM flat mounts (Figure 1). Spectra were grouped by peak emission from the 436 nm excitation. In the CS images, the majority of spectra recovered had relative peak emissions at 510 nm, 530 nm, 570 nm, and 620 nm. These matched spectra previously described1 in the FM images: 510 nm (SDrusen), 530 nm (S1), 580 nm (S2), and 620 nm (S3). Exemplary CS and FM spectra are displayed that peak at 510 nm with 436 nm excitation. Notably, the corresponding spectra at 480 and 505 nm excitation also share common shapes (Figure 2).
Conclusions :
The major fluorophore spectra recovered by the NMF method, including previously described RPE spectra S1, S2, S3, and SDrusen from drusen and subRPE deposits, are shared between CS and FM images at each excitation. These preliminary findings support the robustness of NMF and the concept that these distinct spectra represent distinct fluorophore families. The layered information provided in CS imaging warrants further exploration to complement FM data and inform other imaging and histopathologic analyses.
References
1) Tong et al. Retina 36.Suppl 1 (2016): S127.
2) Mohammed et al. IOVS 60.9 (2019):abstract 3463-3463.
This is a 2020 ARVO Annual Meeting abstract.