Figure 2. Comparison of formulas
Figure 3. Absolute mean errors of previous methods in a cohort study
3
measurements were blinded regarding
algorithm predictions.
Figure 2 displays data from the
study. Predictions based on the three-
dimensional OCT morphology-based
algorithm demonstrated the highest
correlation coefficient with the TLP in
comparison with the Olsen and Haigis
predictions. Bland-Altman plots also
demonstrated tighter 95% limits of
agreement between the predicted lens
position and the TLP. Interestingly,
correlations for TLP predictions were also
higher for the Haigis and Olsen formulas
than previously published results.
Conclusion
Intraoperative three-dimensional OCT
morphology may be superior to current
axial-based formulas in predicting post-
operative ELP. The correlation coeffi-
cients for these formulas in our study us-
ing LACS are also higher than published
values. Further study of this may enable
us to demonstrate the benefit of a more
consistent capsulorhexis.
In addition, TLP may prove to be
useful for outcomes analysis utilizing a
different paradigm. New formulas based
on predicted TLP may help improve
refractive accuracy. For comparison,
instead of a potential 280- to 400-µm
lens position prediction error, compara-
ble to approximately 0.4 D of refraction
error, objective measurement will likely
achieve an error in the order of approxi-
mately 50 µm.
We are currently studying a ray trace
method, as well as a radial basis function
model (similar to that recently proposed
by Dr. Hill), using this data. However,
even with our current, more-basic algo-
rithm, we have demonstrated that we
can obtain better predictive results than
more commonly used methods (Figure
3). We hope this work will ultimately
add to the accuracy and consistency of
postoperative refractive outcomes and
continue to improve the lives of our
patients.
