Abstract
Purpose: :
In vivo histology of corneal structures assessed via the Heidelberg HRT Rostock (HRT) confocal microscope may serve as an indicator of corneal health. However, the resulting amount of data is massive and difficult to manage. The objective of this project was to use image processing techniques to objectively evaluate HRT confocal images for changes in corneal health.
Methods: :
Forty central corneal images in 2um depths were collected in both eyes of 65 subjects with the HRT. A subset of images were selected and manually analyzed by experts to identify epithelial wing cell density and the presence of immune cells. Software was developed to select the best image for wing cells, nerve layer and immune cells. Using selected images, additional software based on advanced image processing techniques was developed to automatically detect and calculate multiple attributes for each cell type. Objective measures from the software were compared to manual annotation to validate the imaging segmentation algorithm.
Results: :
Automatic detection produced better, more consistent and faster results than manual counting for epithelial density and immune cell presence. Software was able to accurately identify objects of interest without human interaction. Wing cells, nerve layer and immune cells were objectively analyzed for density, area, position, shape and consistency in morphology. Software was validated on 390 images and includes a graphical interface which allows any HRT user to easily assess images and export measurements for statistical analysis.
Conclusions: :
This research demonstrates that by using confocal imaging and the associated software described in clinical practice, one can objectively evaluate changes in corneal health, allowing for more timely intervention and appropriate treatment choice. Future development will involve the analysis of additional corneal structures, as well as the use of three dimensional image processing techniques to display 3-D objects and more effectively track changes over time.
Clinical Trial: :
www.clinicaltrials.gov NCT00804999
Keywords: microscopy: confocal/tunneling • cornea: epithelium • image processing