Evaluation of Visual Function in Animal Models

National Eye Institute

Investigators

Linked publications, trials & patents

Abstract

During the past year, we initiated a project, in collaboration with NEI Office of Scientific Information Officer (OISO) and BioTeam, to develop automated segmentation of OCT images for retinal layers using machine-learning (ML) tools. The ML model used was by a U-net convolutional neural network trained using the dice coefficient function as the loss function and the Adam optimizer (learning rate = 0.001), with 665 previously labeled images for training and 166 and 356 images for validation and testing, respectively. To improve the models performance, the label predictions were post-processed by applying a shortest-path algorithm. While we are still improving the model for accuracy and adaption for images from various retinal degenerative animal models, initial release of this program has already been used by several research groups to assist their analysis of OCT images, from comparing retinal layer thickness to extracting metabolic signal of photoreceptors. In addition, we also carried out a study to evaluate fast and slow light-induced changes in murine outer retina OCT as complimentary high spatial resolution functional biomarkers. Fast and slow optical coherence tomography (OCT) responses to light stimulation have been developed to probe outer retinal function with higher spatial resolution than the classical full-field electroretinogram (ERG). However, the relationships of functional information revealed by OCT and ERG are largely unexplored. In this study, we directly compared the fast and slow OCT responses with the ERG. Fast responses i.e., the optoretinogram (ORG) are dominated by reflectance changes in the outer segment (OS) and the inner segment ellipsoid zone (ISez). The ORG OS response has a faster kinetics and higher light sensitivity than the ISez response, and both differ significantly with ERG parameters. Sildenafil-inhibition of phototransduction reduced ORG light sensitivity, suggesting complete phototransduction pathway is needed for ORG responses. Slower OCT responses were dominated for light-induced changes in the external limiting membrane to retinal pigment epithelium (ELM-RPE) thickness and photoreceptor-tip hyporeflective band (HB) magnitudes, with the biggest changes occurring after prolonged light stimulation. Mice with high (129S6/ev) vs. low (C57BL/6J) ATP synthesis efficiency show similar fast ORG and dissimilar slow OCT responses. We propose that ORG reflect passive physiology, such as water movement from photoreceptor, in response to the photocurrent response (measurable by ERG), whereas the slow OCT responses measure mitochondria-driven physiology in the outer retina, such as dark-provoked water removal from the subretinal space.