Purpose:
To predict progression of glaucoma by analyzing retinal nerve fiber layer (RNFL) thickness using the Fast Fourier Analysis.
Methods:
Fifty nine glaucoma patients were followed for a period up to 40 months at the Ophthalmology Department of Semmelweis University. Patients underwent evaluation with white on white perimetry and RNFL measurement using the GDx- Variable Corneal Compensator (GDx-VCC) Scanning Laser Polarimeter approximately every 6 months. The average number of GDX-VCC scans available for the patients examined was 6.54 and the patients were considered progressing if two consecutive visual fields showed progressive damage. The baseline RNFL data (64 sectors) that was available from the GDx-VCC was used in Fast Fourier Analysis and obtaining a classifier. A 10-fold cross validation procedure was performed and a classifier and the classifier that accounted for the maximum amount of variance was used in the study. The classifier obtained from the baseline data was applied to RNFL obtained at different visits and sensitivity, specificity and ROC area was calculated to assess the ability of Fast Fourier Analysis in differentiating progressors from non-progressors at 6, 12 and 18 months prior to progression.
Results:
On basis of visual fields 14 patients showed progressive damage in visual fields which was confirmed on second visit. The mean time to progression from baseline was 22.71 (standard deviation 12.63). The ROC area in identifying progressors from non-progressors was 0.941, 0.828 and 0.826 respectively for 6, 12 and 18 months prior to progression. The sensitivity specificity and ROC and sensitivity at fixed specificity of 90% are given in Table 1.
Conclusions:
The Fast Fourier Analysis can predict progressive glaucomatous damage with moderate accuracy.
Keywords: imaging/image analysis: clinical • image processing • imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound)