Enhancing Histopathological Assessment: Automated Analysis of Bovine and Equine Optic Nerves

The use of specific image analysis algorithms can effectively address the challenges associated with subjective assessment in veterinary pathology. This is particularly important in neuropathology, where key aspects of histopathologic assessment, such as measuring nerve cell loss or the extent of glial proliferation, often depend on the subjective judgment of the operator. To address this issue, we focused our efforts on quantifying nerve fiber density, degree of myelination, amount of interstitial tissue, and glial cell population in normal bovine and equine optic nerves, with the goal of providing a dataset for subsequent evaluation of pathological specimens. Optic nerves from 24 cattle (aged 2 days to 4 years) and 13 horses (aged 5 days to 30 years) regularly slaughtered or euthanized for non-neuro-ophthalmologic conditions were formalin fixed and paraffin embedded. Transverse sections were stained with hematoxylin and eosin, Luxol fast blue (LFB), and Goldner's trichrome. Immunoperoxidase was performed on sections of the optic nerve using glial fibrillary acidic protein (GFAP), oligodendrocyte transcription factor 2 (Olig2), and phosphorylated neurofilaments (2F11) as markers for astrocytes, oligodendrocytes, and axons, respectively. All slides were scanned with the Hamamatsu NanoZoomer and analyzed with AI-driven software (Visiopharm, Hoersholm, Denmark). Optic nerve sections were selected by manually tracing regions of interest (ROIs) on whole slide images. A deep learning trained application then refined the ROIs, excluding interstitial tissue. Multiple sections per slide (ranging from 2 to 5) were analyzed individually within separate ROIs. For LFB-stained slides, another application quantified myelinated fibers by defining a threshold for blue-stained areas relative to the total area analyzed. In Goldner-stained sections, interstitial tissue was distinguished and quantified in a similar manner. 2F11-immunolabeled axons in the optic nerve were also counted separately for each ROI using the threshold method. Myelin sheath/total area (ROI) and interstitial tissue/total area (ROI) calculations were performed on 147 bovine and 129 equine optic nerve sections stained with LFB and Goldner, respectively, while the number of axons in each optic nerve was also counted in an equal number of bovine and equine sections immunolabeled with 2F11. Analysis of GFAP and Olig2 expression completed the dataset. By providing a quantifiable dataset of the optic nerve components, our approach lays the foundation for future evaluation of pathological samples, ultimately improving diagnostic accuracy and research in neurodegenerative diseases.