Comparison of recording systems and analysis methods in specular microscopy.

PURPOSE

To compare corneal endothelial cell images from contact and automated noncontact specular microscopes and to compare endothelial image analysis by the Konan Robo Center Method and the Bio Optics Bambi Corners Method.

METHODS

Twenty-six normal corneas of 13 subjects and 41 penetrating keratoplasties (PKs) of 38 patients were photographed with a Keeler-Konan contact specular microscope and a Konan Noncon Robo automated noncontact specular microscope. (i) After measuring and calibrating the magnification of each instrument, we digitized the cellular apices and analyzed the images from both instruments by using the Corners Method modified to accept x and y calibrations. (ii) Using the internal calibration marks of the Konan Noncon Robo specular microscope for calibration of magnification (as required for the Center Method), we evaluated identical cells on images from this microscope by both the Center Method and the Corners Method. (iii) We evaluated the reproducibility of both methods by repeating measurements on the same image.

RESULTS

(i) When the images were properly calibrated for magnification by using an external scale, endothelial cell density (ECD) of normal corneas was 2,703 +/- 354 (mean +/- SD) cells/mm2 by contact and 2,685 +/- 357 cells/mm2 by noncontact techniques (p = 0.51). ECD of PK corneas was 1,767 +/- 773 cells/mm2 by contact and 1,807 +/- 775 cells/mm2 by noncontact techniques (p = 0.31). (ii) When images from the Konan Noncon Robo specular microscope were calibrated for magnification on the internal marks, the measured ECD from the same noncontact photographs was 6% less (p < 0.001). ECD was then 2,519 +/- 294 cells/mm2 (means +/- SD) by the Center Method and 2,523 +/- 305 cells/mm2 by the Corners Method (p = 0.55) in normal corneas and 1,715 +/- 748 cells/mm2 by the Center Method and 1,731 +/- 763 cells/mm2 by the Corners Method (p = 0.04) in PK corneas. (iii) The coefficient of variation of repeated measurements on the same normal image was 0.0025 for the Centers Method and 0.0099 for the Corners Method.

CONCLUSIONS

(i) Images from the automated noncontact specular microscope may be used interchangeably with those from the contact specular microscope to measure ECD, but only if both are properly calibrated by measuring an external scale. (ii) As a method of analysis, the Center Method is equivalent to the Corners Method in normal corneas, but the proprietary internal calibration of the Center Method, which is required for its use, yields ECDs approximately 6% less than when an external scale is used for distance calibration. (iii) Cell density measurements by both the Center Method and the Corners Method were reproducible within 1%.