The OCT pachymetry maps are highly repeatable: Our study showed a pooled root-mean-square repeatability of 3 μ
m (normal) or 5 μ
m (keratoconus). We applied keratoconus analysis to the central 5-mm area of the OCT pachymetry maps, because we had previously found the repeatability to be better in the 0- to 5-mm diameter zones than in larger diameter zones.39
This was confirmed for keratoconus parameter calculations (I-S and IT-SN) in this study (). Another rationale for applying the analysis to the central 5-mm region is that the cone peak of a keratoconic eye is most likely located inside this region.22
Minimum (or central) corneal thickness is the most frequently cited pachymetric parameter to evaluate keratoconus. Rabinowitz et al,33
Pflugfelder et al,3
and several other research groups29,31,32,34
reported that keratoconic corneas were significantly thinner than normal corneas (). Minimum corneal thickness was the best parameter in our study (highest AROC value, 0.954). However, a low “minimum corneal thickness” value could also be due to generalized corneal thinning, focal thinning, or both; it might not specifically indicate keratoconus. Binder’s study50
showed that a thin cornea may not necessary lead to ectasia after refractive surgeries. Furthermore, the normal corneal thickness range has a wide distribution: The average central corneal thickness of normal eyes is 536±31 μ
m (median ± SD), as determined by a metaanalysis.51
The smaller population SD of 23.7 μ
m found in our study might be due to the relatively small sample size; the 492-μ
m cutoff value derived from our normal reference data might be too high and might result in lower specificity when applied to larger populations. In the future, larger studies will more firmly establish the diagnostic cutoff values and diagnostic performance of OCT pachymetric map analysis.
Average Minimum/Central Corneal Thicknesses of Keratoconic and Normal Eyes Published in the Literature
Focal corneal thinning is possibly a more specific indicator of keratoconus.2
We designed a new parameter—“minimum–median”—to detect focal thinning. The AROC of this difference parameter was not higher than the minimum, but we considered these 2 parameters to be complementary, because they measured different aspects of corneal thinning (focal and general).
In addition to focal thinning, Vinciguerra and Camesasca52
and Binder et al5
summarized that asymmetric and eccentric corneal thinning were also characteristics of keratoconus. To evaluate the asymmetry of the map, we calculated the difference between the average corneal thicknesses from all possible pairs of 4 opposite octants: I-S, IT-SN, T-N, and IN-ST. We found that only the I-S and IT-SN parameters yielded significant differences between the keratoconus and the normal groups (P
<0.0001). The correlation coefficient between I-S and IT-SN was 0.74; therefore, we considered them to have some complementary information and used both octant differences as asymmetry parameters. To evaluate the eccentricity of the corneal thinning, we calculated 2 parameters: the distance from the thinnest corneal point to the corneal vertex, and the vertical location of the thinnest corneal point. We found that the vertical location of the thinnest corneal point had better discriminative power (AROC values, 0.92 vs 0.86), and we therefore selected it as the eccentricity parameter.
Our results suggested that the 5 OCT pachymetric diagnostic parameters we selected were highly complementary. The “1 abnormal” criterion improved the diagnosis power for keratoconus and provided an excellent combination of sensitivity (0.97) and specificity (0.97). And with “2 abnormal” criterion, the specificity of keratoconus diagnosis was very high (1.00). Of course, larger studies are necessary to validate our findings. Moreover, our study is a preliminary study consisting only of keratoconus cases that had been confirmed by clinical (slit lamp) signs. Therefore, the sensitivity of OCT in the detection of forme fruste keratoconus (those without clinical signs) has yet to be established.
An OCT pachymetry map analysis can provide decisive diagnostic information in cases where topography is ambiguous. In case 1, part of the topography map was loss might owing to poor corneal reflex. In contrast, OCT was able to provide pachymetry map over a wide area of central cornea. In cases 2 and 3, KISA% analysis identified the topography maps as within normal range, because asymmetry or skew in the bowtie astigmatism pattern were masked by irregularities that might be due to a lid artifact or tear film disturbance. Performance of OCT was not affected by these conditions. It is an important advantage that OCT can reliably map the corneal thickness of normal, postoperative, and opacified corneas.39-41
One limitation of the current OCT technology is that interpolation is used in the central 0.5- to 1.0-mm diameter and among the 8 radial scans. Thus, small areas of corneal thickness variation might be missed.39
Another limitation of the OCT system used in this study was that it was not fast enough to measure the corneal topography; therefore, it could not completely replace a Placido-ring topography system. In a previous study, we found the repeatability of anterior corneal power by this OCT system to be 0.79 D,53
significantly worse than the repeatability of standard keratometry (0.2 D) or Placido-ring topography (0.1 D).54
These findings suggested that the scan speed of 2000 a-scans/sec might not be sufficient to overcome the motion error for surface contour measurements. Thickness measurement was much more robust, because axial eye motion moved the anterior and posterior surface equally and had little effect on the thickness measurement. Therefore, we concentrated on pachymetry map analysis in this study. However, the speed limitation is not intrinsic to OCT; new Fourier-domain OCT technology is capable of >10-fold higher scan speeds.55
We are currently evaluating Fourier-domain OCT as a method to map both pachymetry and topography.
A commercially available OCT instrument similar in performance (wavelength, scan range, speed, signal) to the anterior segment OCT prototype used in this study, the Visante anterior segment OCT system (Carl Zeiss Meditec Inc.), was approved by the US Food and Drug Administration in October 2005. Thus, the keratoconus screening method presented in this study can now be applied by other researchers and clinicians. Because the Visante software uses different image processing algorithms and has a slightly different parameter output format, we are performing another study to confirm our findings using the commercial unit.
In summary, we showed that OCT pachymetry map-based analysis could detect abnormal corneal thinning in keratoconus eyes. We provide a set of 5 diagnostic parameters and cutoff values that performed well together in our small study.