Abstract
Purpose: :
We aimed to evaluate automated and manual optical coherence tomography (OCT) registration across visits in eyes with neovascular age-related macular degeneration (AMD) receiving treatment in order to determine the rate of successful automatic registration, rate of successful manual registration in cases of failed automatic registration, and quality of successful registration. We also compared thickness values of manually registered scans to unregistered scans in cases of failed automatic registration.
Methods: :
After IRB approval, Cirrus SD-OCT (Carl Zeiss Meditec Inc., Dublin, CA) scans from 100 eyes with neovascular AMD were obtained at 3 different visits including baseline, 1 month after an anti-vascular endothelial growth factor (VEGF) drug injection, and a third visit some time later. Automatic registration of each follow-up scan to the baseline scan was either successful or failed. Among cases of failed automated registration, manual registration was attempted. All cases were graded for registration quality. For each case with failed automatic but successful manual registration, difference in percent thickness change between the unregistered and registered macular change map was calculated for each subfield. A difference greater than 10% was considered clinically relevant.
Results: :
142 (70%) of 200 paired images were registered automatically. Among the 57 (29%) OCT pairs which failed automatic registration, manual registration was successful in 55. Registration quality grades in cases of successful automatic registration were perfect or good in 55%, acceptable in 36%, and poor in 9%. Among the 55 cases which required manual registration, the quality grades were perfect or good in 38%, acceptable in 47%, and poor in 15%. The difference between "unregistered" and "manually registered" percent thickness change calculated for the central subfield was greater than 10% in 10 (18%) of the 55 cases that required manual registration. Analysis of the inner and outer subfields showed a clinically relevant difference in 11 (20%) and 3 (5.4%) cases, respectively.
Conclusions: :
Our results suggest that automatic OCT registration succeeds in 70% (95% confidence interval [CI]: 65% to 78%) of paired images with neovascular AMD undergoing treatment. Subsequent manual registration appears to succeed in almost all which fail automatic registration. This information may be important when OCT guides retreatment decisions in eyes with neovascular AMD.
Keywords: imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • age-related macular degeneration