Specular Microscopy Ancillary Study Methods for Donor Endothelial Cell Density Determination of Cornea Donor Study Images

• Published in: Current eye research Vol. 31; no. 4; pp. 319 - 327
• Main Authors: Benetz, Beth Ann, Gal, Robin L., Ruedy, Katrina J., Rice, Carmella, Beck, Roy W, Kalajian, Andrea D., Lass, Jonathan H.
• Format: Journal Article
• Language: English
• Published: England Informa UK Ltd 01-04-2006; Taylor & Francis
• Subjects: Cell Count; cell density; endothelium; Endothelium, Corneal - cytology; Eye Banks; Humans; Microscopy - methods; Observer Variation; Ophthalmology - methods; Quality Control; Reproducibility of Results; specular microscopy; Tissue Donors
• ISSN: 0271-3683; 1460-2202
• DOI: 10.1080/02713680500536738

Purpose: To describe reliable methods for determining central corneal endothelial cell density (ECD) in a multicenter eye bank study. Methods: The Specular Microscopy Reading Center utilized a dual-grading procedure and adjudication process to classify image quality and determine ECD for a subset of donor endothelial images obtained in the Specular Microscopy Ancillary Study, which is part of the Cornea Donor Study.1 Two certified readers classified images as analyzable (excellent, good, fair) or unanalyzable and determined the ECD using a variable frame technique. An adjudicator also evaluated the images if quality classifications by the two readers differed by one grade, if any reader found the image unanalyzable, and/or if the ECD determination between the two readers was ≥ 5%. Results: Image quality categorization by the two readers was identical for 441 (64%) of 688 donor images. The ECD differed by < 5% for 442 (69%) of the 645 analyzable images. The ECD determined by the adjudicator was < 5% different than the ECD determined by at least one reader for 193 (95%) of the 203 remaining images. Conclusions: The dual-grading and adjudication procedures produce reliable, reproducible assessments of image quality and ECD. The importance of two independent readings is evident in that image quality ratings differed between the two readers by one grade in 36% of all images and ECD counts differed by ≥5% for 31% of analyzable images.