Automated Quantification of Photoreceptor Outer Segments in Developing and Degenerating Retinas on Microscopic Images across Scales

Abstract

Purpose : The functionality of photoreceptors is highly dependent on outer segments (OS). These membranous structures generate biochemical signals from incident light. While OS formation and degeneration is qualitatively assessed on microscopic images, reliable methodology for quantitative analyses is still limited. Here, we developed quantification methods to evaluate OS maturation and quality using automated image analysis.

Methods : OS formation was examined during development and in adulthood (between 4 days and 35 weeks of age) of C57Bl/6JRj wild-type mice via light and transmission electron microscopy (TEM).
To quantify number, size, shape and fluorescence intensity of OS, retinal cryo-sections were immuno-stained for s-opsin. Fluorescence images were used to train a robust classifier based on supervised machine learning for automated image segmentation. Characteristic features of resulting labels were extracted to quantify maturation of cone OSs. Subsequently, this quantification method was applied to characterize OS degeneration in ‘cone photoreceptor function loss 1’ (Cpfl1) mice between 8 days and 35 weeks of age.
TEM images were used to establish an ultrastructural quantification measurement for alignment of OS membrane stacks. Images were analyzed using a custom written MATLAB code to extract the orientation of membranes from the image gradient and their alignment (coherency). We used this analysis to quantify the OS morphology of wild-type and two inherited retinal degeneration (‘retinal degeneration 19’ (rd19) and ‘rhodopsin knock-out’ (RhoKO)) mouse lines at 1 month of age.

Results : Both automated analysis technologies provided robust characterization and quantification of OS based on fluorescence microscopy or TEM images. Automated assessments showed an increase in OS number, volume and membrane coherency during wild-type postnatal development, while a decrease in all three observables was detected in different retinal degeneration mouse models.

Conclusions : Automated image segmentation by a classifier based on machine learning and analysis of the orientation of membrane stacks using fluorescent or TEM images, respectively, allow quantitative evaluation of photoreceptor OS formation and quality. These approaches are useful for in depth analysis of OS in developmental studies, for disease modeling or after therapeutic interventions affecting photoreceptors.

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