The present study aimed to elucidate if comparison of angular segments of Pigment epithelium central limit- Inner limit
of the retina Minimal Distance, measured over 2π radians in the frontal plane (PIMD-2π) between visits of a patient,
renders sufficient precision for detection of loss of nerve fibers in the optic nerve head. An optic nerve head raster
scanned cube was captured with a TOPCON 3D OCT 2000 (Topcon, Japan) device in one early to moderate stage
glaucoma eye of each of 13 patients. All eyes were recorded at two visits less than 1 month apart. At each visit,
3 volumes were captured. Each volume was extracted from the OCT device for analysis. Then, angular PIMD was
segmented three times over 2π radians in the frontal plane, resolved with a semi-automatic algorithm in 500 equally
separated steps, PIMD-2π. It was found that individual segmentations within volumes, within visits, within subjects can
be phase adjusted to each other in the frontal plane using cross-correlation. Cross correlation was also used to phase
adjust volumes within visits within subjects and visits to each other within subjects. Then, PIMD-2π for each subject was
split into 250 bundles of 2 adjacent PIMDs. Finally, the sources of variation for estimates of segments of PIMD-2π were
derived with analysis of variance assuming a mixed model. The variation among adjacent PIMDS was found very small
in relation to the variation among segmentations. The variation among visits was found insignificant in relation to the
variation among volumes and the variance for segmentations was found to be on the order of 20 % of that for volumes.
The estimated variances imply that, if 3 segmentations are averaged within a volume and at least 10 volumes are
averaged within a visit, it is possible to estimate around a 10 % reduction of a PIMD-2π segment from baseline to a
subsequent visit as significant. Considering a loss rate for a PIMD-2π segment of 23 μm/yr., 4 visits per year, and
averaging 3 segmentations per volume and 3 volumes per visit, a significant reduction from baseline can be detected
with a power of 80 % in about 18 months. At higher loss rate for a PIMD-2π segment, a significant difference from
baseline can be detected earlier. Averaging over more volumes per visit considerably decreases the time for detection of
a significant reduction of a segment of PIMD-2π. Increasing the number of segmentations averaged per visit only
slightly reduces the time for detection of a significant reduction. It is concluded that phase adjustment in the frontal plane
with cross correlation allows high precision estimates of a segment of PIMD-2π that imply substantially shorter followup
time for detection of a significant change than mean deviation (MD) in a visual field estimated with the Humphrey
perimeter or neural rim area (NRA) estimated with the Heidelberg retinal tomograph.
|