Background: Regarding the growing interest and importance of understanding the cellular changes of the cornea in diseases, a quantitative cellular characterization of the epithelium is becoming increasingly important. Towards this, the latest research offers considerable improvements in imaging of the cornea by confocal laser scanning microscopy (CLSM). This study presents a pipeline to generate normative morphological data of epithelial cell layers of healthy human corneas.
Methods: 3D in vivo CLSM was performed on the eyes of volunteers (n=25) with a Heidelberg Retina Tomograph II equipped with an in-house modified version of the Rostock Cornea Module implementing two dedicated piezo actuators and a concave contact cap. Image data were acquired with nearly isotropic voxel resolution. After image registration, stacks of en-face sections were used to generate full-thickness volume data sets of the epithelium. Beyond that, an image analysis algorithm quantified en-face sections of epithelial cells regarding the depth-dependent mean of cell density, area, diameter, aggregation (Clark and Evans index of aggregation), neighbor count and polygonality.
Results: Imaging and cell segmentation were successfully performed in all subjects. Thereby intermediated cells were efficiently recognized by the segmentation algorithm while efficiency for superficial and basal cells was reduced. Morphological parameters showed an increased mean cell density, decreased mean cell area and mean diameter from anterior to posterior (5,197.02 to 8,190.39 cells/mm²; 160.51 to 90.29 µm²; 15.9 to 12.3 µm respectively). Aggregation gradually increased from anterior to posterior ranging from 1.45 to 1.53. Average neighbor count increased from 5.50 to a maximum of 5.66 followed by a gradual decrease to 5.45 within the normalized depth from anterior to posterior. Polygonality gradually decreased ranging from 4.93 to 4.64 sides of cells. The neighbor count and polygonality parameters exhibited profound depth-dependent changes.
Conclusions: This in vivo study demonstrates the successful implementation of a CLSM-based imaging pipeline for cellular characterization of the human corneal epithelium. The dedicated hardware in combination with an adapted image registration method to correct the remaining motion-induced image distortions followed by a dedicated algorithm to calculate characteristic quantities of different epithelial cell layers enabled the generation of normative data. Further significant effort is necessary to improve the algorithm for superficial and basal cell segmentation.