ESCRS - FP16.10 - An Innovative Specular Microscopy Software Application To Evaluate Light Microscopy Images To Obtain Quantitative Corneal Endothelium Data

An Innovative Specular Microscopy Software Application To Evaluate Light Microscopy Images To Obtain Quantitative Corneal Endothelium Data

Published 2025 - 43rd Congress of the ESCRS

Reference: FP16.10 | Type: Free paper

Purpose

To develop and characterize a specular microscope software-assisted method to quantitively evaluate corneal endothelial cell density (ECD) and morphological parameters using light microscopy images. To evaluate agreement of the new method with the evaluation techniques routinely performed at the eye banks as light microscopy and specular microscopy.

Setting

The study was conducted on human corneas unsuitable for transplantation, preserved in hypothermic conditions. A total of 283 endothelial images from 31 corneas were analyzed using standard light microscopy and specular microscopy techniques. To assess analysis success rates, 311 images from 70 corneas stored at 2–8°C between 2023 and 2024 were examined. The new method was applied by processing light microscopy images with a specular microscopy-dedicated software.

Methods

A total of 283 images from endothelia of 31 corneas unsuitable for transplantation in hypothermic storage were obtained and analysed with standard light microscopy and specular microscopy. The same light microscopy images were processed with a dedicated specular microscopy software to determine ECD, CV% and HEX%. The agreement between the three methods between different operators investigated with Bland-Altman statistical analysis. Additionally, the success rates of light microscopy and specular microscopy in generating analyzable corneal endothelial images were evaluated in 311 assessments conducted on 70 corneas stored at 2–8°C between 2023 and 2024.

Results

The new method aligned with specular microscopy in ECD measurement, while light microscopy showed a significant difference from both. HEX% values matched specular microscopy, but CV% varied significantly. Interoperator analysis revealed discrepancies in ECD with light microscopy, whereas the new method ensured consistency for ECD and HEX%, though CV% showed bias. Despite statistical significance, differences in ECD, CV%, and HEX% were clinically irrelevant. Light microscopy produced analysable images in 96.8% of cases, whereas specular microscopy in the 72.2% of cases. The new method, based on light microscopy images, maintained this success rate while enabling quantitative morphological analysis.

Conclusions

The new method utilizes images obtained from light microscopy and allows to obtain ECD as well as endothelial morphological data CV% and HEX%, which had previously been limited only to specular microscopy. The new method also showed statistical agreement in ECD and HEX% determination with specular microscopy, while a statistically significant but clinically acceptable bias was identified in ECD determination between light microscopy and both the new method and specular microscopy, as well as in CV% between the new method and specular microscopy.

Additionally, the new method exhibited lower interoperator variability compared to light microscopy and a higher success rate than specular microscopy.