Comparison of clinician judgment with digitized image analysis in the detection of induced optic disk change in...

To compare the ability of clinicians to detect change in the photographic appearance of the optic disk with the performance of a system for digitized image analysis.

In 11 monkey eyes, a Topcon Imagenet System was used to acquire eight digitized image pairs and four stereoscopic photographs at an intraocular pressure of 10 mm Hg, and then, again, 45 minutes after intraocular pressure was increased to 45 mm Hg. We recently reported detection of global (ten of 11 eyes) and regional (11 of 11 eyes) change in the digitized images of these eyes by using two new statistical strategies for optic disk analysis. For the current study, we evaluated the ability of three clinicians (the authors) to detect a change within the stereoscopic photographs of these 11 optic disks. For each eye, the eight stereoscopic photographs (four at intraocular pressure of 10 mm Hg and four at intraocular pressure of 45 mm Hg) were developed as stereoscopic slides and arranged into four pairs (10/10, 45/45, 10/45, and 45/10 mm Hg). Thus, two pairs represented no change in intraocular pressure (10/10 and 45/45 mm Hg) and the other two pairs represented either an increase or a decrease in intraocular pressure (10/45 and 45/10 mm Hg). The 44 pairs of stereoscopic slides (four pairs for each of 11 eyes) were masked then randomly mixed. On two separate occasions, each clinician evaluated each pair of stereoscopic slides for the presence of absence of optic disk change.

Reproducibility between the two readings of each clinician ranged from .50 to .64 (kappa statistic). Clinicians correctly detected change (as detected by image analysis) within 45% to 64% of the 10/45 and 45/10 pairs of stereoscopic slides. Clinicians correctly indentified no change within 86% to 100% of the 10/10- and 45/45-mm Hg pairs of stereoscopic slides. Clinicians correctly identified no change significantly more often than change (P < .01, chi 2 test). Change was not detected consistently by all three clinicians in any of the 11 eyes.

In a controlled experimental setting, digitized image acquisition with extensive secondary statistical analysis more sensitively detected small short-term changes in the surface of the optic disks of monkeys than did three masked clinicians.