Abstract
Purpose: :
(1) To determine whether use of shape–based retinal nerve fiber layer (RNFL) thickness analysis methods (Wavelet–Fourier Analysis [WFA] and Fast–Fourier Analysis [FFA]) is more beneficial for the Optical Coherence Tomograph (OCT) or the GDx–VCC polarimeter for detecting glaucoma. (2) To compare the performance of the two devices for: (a) their standard methods of glaucoma detection (Inferior Average and Nerve Fiber Indicator, NFI) and (b) when shape–based analysis is used.
Methods: :
RNFL estimates were obtained at L.V. Prasad Eye Institute using OCT (OCT–3 Stratus, Carl Zeiss Meditec Inc.) and GDx–VCC (Carl Zeiss Meditec Inc.) from 136 eyes of 136 individuals (73 healthy and 63 early glaucoma, as defined by standard automated perimetry [average MD –1.72, SD=1.45 and –3.69, SD=1.61 respectively]). TSNIT VCC data from rings of standard ("small" ring 27–35 pixel) and custom (33–41 pixel) sizes were used and compared to TSNIT OCT data from the standard 3.46 mm circle (37 equivalent pixel). Independent training and test samples were obtained using 10–fold cross–validation. WFA and FFA with asymmetry were performed on the RNFL TSNIT estimates and linear discriminant functions obtained (Fisher). Performance of WFA, FFA, and the standard metrics of OCT (Inferior Average) and GDX–VCC (NFI) were evaluated by calculating sensitivity at 90% specificity and area under the ROC curve (AUC).
Results: :
In general, (1) OCT better detected glaucoma than GDx–VCC and (2) both shape–based methods detected glaucoma better than the manufacturers’ standard metrics for both devices. With OCT data, AUC was 0.935 & 0.911 for WFA and FFA, respectively, and 0.852 for Inferior Average. With GDx–VCC data, AUC was 0.885 & 0.913 for WFA and FFA, respectively and 0.833 for NFI. The sensitivity at 90% specificity obtained using OCT RNFL estimates for WFA, FFA and Inferior Average were 0.873, 0.794 and 0.656 respectively. The sensitivity and specificity obtained using GDx–VCC RNFL estimates for WFA, FFA and the NFI were 0.762, 0.767 and 0.667 respectively. GDx–VCC performance with WFA or FFA is not clearly affected at the custom radius (33–41 pixels).
Conclusions: :
Performance of both OCT–3 and GDx–VCC devices are improved by shape–based analysis methods. In the present sample OCT–3 was better than GDx–VCC at detecting glaucoma with standard measures or WFA but not with FFA.
Keywords: imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • imaging/image analysis: clinical • visual fields