A Novel Hybrid Software-Assisted Method to Evaluate Quantitatively Corneal Endothelium From Light Microscopy Images

Light (LM) and specular microscopies (SM) are standard techniques used by eye banks during corneal endothelial cell density (ECD) and morphology evaluation. This study aimed to develop a novel Hybrid method (HY) that integrates the benefits of both SM and LM while minimizing their drawbacks. A total of 283 endothelial images from LM and SM were analyzed from 31 corneas. For HY analysis, LM images were processed using SM-dedicated software to semi-automatically determine ECD, CV% (coefficient of variation), and HEX% (hexagonality). Agreements between LM, SM, and HY, as well as inter-operator bias, were assessed using the Bland-Altman analysis. Evaluability of corneas with LM, SM, and HY was recorded during 311 examinations on 70 corneas. HY agreed with SM in ECD determination, while LM differed from HY (bias: 134 cells/mm²) and SM (bias: 115 cells/mm²). HY showed agreement with SM in HEX% determination, while a bias of 3.4% was observed in CV%. Inter-operator variability analysis showed significant differences in LM evaluations (ECD, EC morphology score). For HY, no significant inter-operator bias was obtained in ECD and HEX%, whereas CV% displayed a significant bias (3.1%). Corneal evaluability was significantly higher in LM and HY (both 96.5%) than in SM (72.7%). HY enabled quantitative ECD and morphology investigation of corneal endothelia using LM-obtained images. HY, SM, and LM techniques statistically agreed in ECD and morphology examinations, or showed clinically acceptable bias. The HY method demonstrated lower inter-operator variability than LM and higher evaluability than SM.