AxonDeep: Automated Optic Nerve Axon Segmentation in Mice With Deep Learning

Mitochondrially Targeted Gene Therapy Rescues Visual Loss in a Mouse Model of Leber's Hereditary Optic Neuropathy

Leber's hereditary optic neuropathy (LHON) is a common mitochondrial genetic disease, causing irreversible blindness in young individuals. Current treatments are inadequate, and there is no definitive cure. This study evaluates the effectiveness of delivering wildtype human NADH ubiquinone oxidoreductase subunit 4 (hND4) gene using mito-targeted AAV(MTSAAV) to rescue LHOH mice. We observed a declining pattern in electroretinograms amplitudes as mice aged across all groups (p < 0.001), with significant differences among groups (p = 0.023; Control vs. LHON, p = 0.008; Control vs. Rescue, p = 0.228). Inner retinal thickness and intraocular pressure did not change significantly with age or groups. Compared to LHON mice, those rescued with wildtype hND4 exhibited improved retinal visual acuity (0.29 ± 0.1 cy/deg vs. 0.15 ± 0.1 cy/deg) and increased functional hyperemia response (effect of flicker, p < 0.001, effect of Group, p = 0.004; Interaction Flicker × Group, p < 0.001). Postmortem analysis shows a marked reduction in retinal ganglion cell density in the LHON group compared to the other groups (Effect of Group, p < 0.001, Control vs. LHON, p < 0.001, Control vs. Rescue, p = 0.106). These results suggest that MTSAAV-delivered wildtype hND4 gene rescues, at least in part, visual impairment in an LHON mouse model and has the therapeutic potential to treat this disease.

AxonQuantifier: A semi-automated program for quantifying axonal density from whole-mounted optic nerves

BACKGROUND

Here, we present a semi-automated method for quantifying retinal ganglion cell (RGC) axon density at different distances from the optic nerve crush site using longitudinal, confocal microscopy images taken from whole-mounted optic nerves. This method employs the algorithm AxonQuantifier which operates on the freely available program, ImageJ.

NEW METHOD

To validate this method, seven adult male Long Evans rats underwent optic nerve crush injury followed by in vivo treatment with electric fields of varying strengths for 30 days to produce optic nerves with a wide range of axon densities distal to the optic nerve crush site. Prior to euthanasia, RGC axons were labelled with intravitreal injections of cholera toxin B conjugated to Alexa Fluor 647. After dissection, optic nerves underwent tissue clearing, were whole-mounted, and imaged longitudinally using confocal microscopy.

COMPARISON WITH EXISTING METHODS

Five masked raters quantified RGC axon density at 250, 500, 750, 1000, 1250, 1500, 1750, and 2000 µm distances past the optic nerve crush site for the seven optic nerves manually and using AxonQuantifier. Agreement between these methods was assessed using Bland-Altman plots and linear regression. Inter-rater agreement was assessed using the intra-class coefficient.

RESULTS

Semi-automated quantification of RGC axon density demonstrated improved inter-rater agreement and reduced bias values as compared to manual quantification, while also increasing time efficiency 4-fold. Relative to manual quantification, AxonQuantifier tended to underestimate axon density.

CONCLUSIONS

AxonQuantifier is a reliable and efficient method for quantifying axon density from whole mount optic nerves.

RGC-Net: An Automatic Reconstruction and Quantification Algorithm for Retinal Ganglion Cells Based on Deep Learning

Purpose

The purpose of this study was to develop a deep learning-based fully automated reconstruction and quantification algorithm which automatically delineates the neurites and somas of retinal ganglion cells (RGCs).

Methods

We trained a deep learning-based multi-task image segmentation model, RGC-Net, that automatically segments the neurites and somas in RGC images. A total of 166 RGC scans with manual annotations from human experts were used to develop this model, whereas 132 scans were used for training, and the remaining 34 scans were reserved as testing data. Post-processing techniques removed speckles or dead cells in soma segmentation results to further improve the robustness of the model. Quantification analyses were also conducted to compare five different metrics obtained by our automated algorithm and manual annotations.

Results

Quantitatively, our segmentation model achieves average foreground accuracy, background accuracy, overall accuracy, and dice similarity coefficient of 0.692, 0.999, 0.997, and 0.691 for the neurite segmentation task, and 0.865, 0.999, 0.997, and 0.850 for the soma segmentation task, respectively.

Conclusions

The experimental results demonstrate that RGC-Net can accurately and reliably reconstruct neurites and somas in RGC images. We also demonstrate our algorithm is comparable to human manually curated annotations in quantification analyses.

Translational Relevance

Our deep learning model provides a new tool that can trace and analyze the RGC neurites and somas efficiently and faster than manual analysis.

AxoNet 2.0: A Deep Learning-Based Tool for Morphometric Analysis of Retinal Ganglion Cell Axons

Purpose

Assessment of glaucomatous damage in animal models is facilitated by rapid and accurate quantification of retinal ganglion cell (RGC) axonal loss and morphologic change. However, manual assessment is extremely time- and labor-intensive. Here, we developed AxoNet 2.0, an automated deep learning (DL) tool that (i) counts normal-appearing RGC axons and (ii) quantifies their morphometry from light micrographs.

Methods

A DL algorithm was trained to segment the axoplasm and myelin sheath of normal-appearing axons using manually-annotated rat optic nerve (ON) cross-sectional micrographs. Performance was quantified by various metrics (e.g., soft-Dice coefficient between predicted and ground-truth segmentations). We also quantified axon counts, axon density, and axon size distributions between hypertensive and control eyes and compared to literature reports.

Results

AxoNet 2.0 performed very well when compared to manual annotations of rat ON (R2 = 0.92 for automated vs. manual counts, soft-Dice coefficient = 0.81 ± 0.02, mean absolute percentage error in axonal morphometric outcomes < 15%). AxoNet 2.0 also showed promise for generalization, performing well on other animal models (R2 = 0.97 between automated versus manual counts for mice and 0.98 for non-human primates). As expected, the algorithm detected decreased in axon density in hypertensive rat eyes (P ≪ 0.001) with preferential loss of large axons (P < 0.001).

Conclusions

AxoNet 2.0 provides a fast and nonsubjective tool to quantify both RGC axon counts and morphological features, thus assisting with assessing axonal damage in animal models of glaucomatous optic neuropathy.

Translational Relevance

This deep learning approach will increase rigor of basic science studies designed to investigate RGC axon protection and regeneration.

Reverse translation of artificial intelligence in glaucoma: Connecting basic science with clinical applications

Artificial intelligence (AI) has been approved for biomedical research in diverse areas from bedside clinical studies to benchtop basic scientific research. For ophthalmic research, in particular glaucoma, AI applications are rapidly growing for potential clinical translation given the vast data available and the introduction of federated learning. Conversely, AI for basic science remains limited despite its useful power in providing mechanistic insight. In this perspective, we discuss recent progress, opportunities, and challenges in the application of AI in glaucoma for scientific discoveries. Specifically, we focus on the research paradigm of reverse translation, in which clinical data are first used for patient-centered hypothesis generation followed by transitioning into basic science studies for hypothesis validation. We elaborate on several distinctive areas of research opportunities for reverse translation of AI in glaucoma including disease risk and progression prediction, pathology characterization, and sub-phenotype identification. We conclude with current challenges and future opportunities for AI research in basic science for glaucoma such as inter-species diversity, AI model generalizability and explainability, as well as AI applications using advanced ocular imaging and genomic data.

Deep learning and optical coherence tomography in glaucoma: Bridging the diagnostic gap on structural imaging

Glaucoma is a leading cause of progressive blindness and visual impairment worldwide. Microstructural evidence of glaucomatous damage to the optic nerve head and associated tissues can be visualized using optical coherence tomography (OCT). In recent years, development of novel deep learning (DL) algorithms has led to innovative advances and improvements in automated detection of glaucomatous damage and progression on OCT imaging. DL algorithms have also been trained utilizing OCT data to improve detection of glaucomatous damage on fundus photography, thus improving the potential utility of color photos which can be more easily collected in a wider range of clinical and screening settings. This review highlights ten years of contributions to glaucoma detection through advances in deep learning models trained utilizing OCT structural data and posits future directions for translation of these discoveries into the field of aging and the basic sciences.

Biological Correlations and Confounders for Quantification of Retinal Ganglion Cells by Optical Coherence Tomography Based on Studies of Outbred Mice

Purpose

Despite popularity of optical coherence tomography (OCT) in glaucoma studies, it's unclear how well OCT-derived metrics compare to traditional measures of retinal ganglion cell (RGC) abundance. Here, Diversity Outbred (J:DO) mice are used to directly compare ganglion cell complex (GCC) thickness measured by OCT to metrics of retinal anatomy measured ex vivo with retinal wholemounts and optic nerve histology.

Methods

J:DO mice (n = 48) underwent fundoscopic and OCT examinations, with automated segmentation of GCC thickness. RGC axons were quantified from para-phenylenediamine-stained optic nerve cross-sections and somas from BRN3A-immunolabeled retinal wholemounts, with total inner retinal cellularity assessed by TO-PRO and subsequent hematoxylin staining.

Results

J:DO tissues lacked overt disease. GCC thickness, RGC abundance, and total cell abundance varied broadly across individuals. GCC thickness correlated significantly to RGC somal density (r = 0.58) and axon number (r = 0.44), but not total cell density. Retinal area and nerve cross-sectional area varied widely. No metrics were significantly influenced by sex. In bilateral comparisons, GCC thickness (r = 0.95), axon (r = 0.72), and total cell density (r = 0.47) correlated significantly within individuals.

Conclusions

Amongst outbred mice, OCT-derived measurements of GCC thickness correlate significantly to RGC somal and axon abundance. Factors limiting correlation are likely both biological and methodological, including differences in retinal area that distort sampling-based estimates of RGC abundance.

Translational Relevance

There are significant-but imperfect-correlations between GCC thickness and RGC abundance across genetic contexts in mice, highlighting valid uses and ongoing challenges for meaningful use of OCT-derived metrics.