Zeiss CIRRUS HD-OCT 500, Руководство пользователя

Страница: 236 / 352

Analysis Related Report Options
2660021169012 Rev. A 2017-12 CIRRUS HD-OCT User Manual
10-14
Layer Thickness Measurements
” – M.R. Horne, T. Callan, M. Durbin, T. Abunto; Poster
4624, ARVO 2008).
The difference between a current exam and the Baseline is assumed to have a normal
distribution with a standard deviation determined from clinical measurements on subjects
over a short period of time. Only 5% of paired measurements are expected to have an
absolute difference more than 1.96 times the standard deviation of differences observed in
an in–house reproducibility study of normals, which is the test–retest variability. This is
also equal to 2.77 times the reproducibility standard deviation observed.
Figure 10-11
illustrates a normal distribution of differences, centered at a mean difference of zero. The
yellow line shows the cutoff for declaring ”Possible Loss” or ”Likely Loss” and the lavender
line shows the cutoff for declaring ”Possible Increase.” For any individual comparison of a
measurement to Baseline, only 2.5% of measurements are expected to show change to the
left of the yellow line when real loss has not occurred, and only 2.5% of measurements are
expected to show change to the right of the lavender line.
Because thickness maps and profiles have multiple points available for testing, the
observed rate of false positives would be higher than 5% if the cutoff is set based on the
95% confidence limit depicted in
Figure 10-11 . To achieve a reasonable false positive
rate of no more than 5% for any given visit, the limits are set at 99% for the TSNIT profile
and 99.5% for the change map.
Using Confirmation to Improve Specificity
In order to increase the specificity of the measurement over multiple visits, CIRRUS also
requires that statistically significant change from Baseline be observed for at least two
pairs of measurements when only three measurements are available, and for at least three
pairs of measurements when four or more measurements are available. In this case,
CIRRUS will report ”Possible Loss” for a parameter. For example, a super-pixel on the
change map will be colored yellow, or the RNFL Profile will show a yellow region between
the Baseline and current scans, or the Average Thickness plot will show a yellow symbol for
that visit. If, on the following visit, these same conditions are met for the same parameter,
CIRRUS will report ”Likely Loss”, because now the change has been flagged for more than
one visit. These confirmation strategies help improve the specificity, and reduce the effects
of individual outlier measurements.
NOTE: RNFL thickness is expected to decrease slowly as a function of normal aging. The
RNFL data collected for the normative database (
"RNFL and ONH Normative
Databases" on page A-15
) showed a rate of loss for overall thickness of –0.2 μ m per
year, with a 95% confidence interval of –0.25 to 0.13 μ m per year. For superior thickness,
the rate was –0.25 μ m per year (–0.35, –0.15), and for inferior thickness, it was –0.3 μ m
per year (–0.42, –0.21). This slow rate of change is consistent with observations of RNFL
thickness loss measured on Stratus OCT. All of these results are based on cross–sectional
data, and an individual patient’s normal aging rate may vary. Because the exact rate of
change for any individual is unknown, CIRRUS reports the 95% confidence limits on the
slope. In the usual statistical understanding, a rate of change would be considered