Improved fluorescence retinal image quality with phase diversity

Abstract

Purpose : Optimized fluorescence image quality could better reveal cellular and subcellular structures in the living eye, but such images are degraded by residual aberrations even with confocal adaptive optics. To address this problem, we explored the value of phase diversity (PD). PD is a computational imaging method that, counterintuitively, can use a defocused image to better estimate the effect of the optics alone on an in-focus image, allowing deconvolution to more effectively remove blur from the retinal structure of interest. Here we demonstrate the resolution and contrast improvement of fluorescent images of dendritic fields of retinal ganglion cells (RGCs) in the living macaque eye provided by PD.

Methods : Intravitreal injections of AAV2-CaMKII-mCherry were made in one macaque and the resulting labeled RGCs were imaged with a confocal adaptive optics scanning light ophthalmoscope (AOSLO) (90 uW excitation @ 561 nm, emission @ 607/40 nm). CaMKII labels only rare RGCs, reducing obscuration from overlapping dendritic arbors. For PD processing, two consecutive images of the same retinal location were obtained by summing 600 frames for each. One image was in focus and a second image was defocused by 0.1 diopter.

Results : Dendritic structures of rare RGCs were labeled and successfully imaged with CaMKII in the living eye. PD provided an appreciable resolution improvement compared with confocal AO imaging alone. The cross-sections of 20 dendrites across multiple locations show an average 18% ± 6% (SD) decrease in FWHM width. More importantly, the combination of lower background noise and enhanced signal intensity increased the signal-to-noise ratio by a factor of 3, allowing more dendritic structures to be seen.

Conclusions : PD can improve AOSLO resolution and contrast by reducing blur from residual aberrations, enabling better visualization of structures at a fine spatial scale. Such morphological information combined with in vivo functional calcium imaging could improve the accuracy of classifying retinal neurons without the need to excise the retina. In the future, more selective labeling than provided by CaMKII could aid in tracing the dendrites from individual cells. Moreover, PD is compatible with other approaches such as image scanning microscopy.

This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.