Comparison of Shape-based Analysis of Retinal Nerve Fiber Layer Data Obtained From OCT and GDx-VCC

PURPOSE: To directly compare in 1 population: (1) the performance of Optical Coherence Tomograph (OCT) and GDx-Variable Corneal Compensator (VCC) when using Wavelet-Fourier Analysis (WFA) and Fast-Fourier Analysis (FFA), (2) the performance of these shape-based and standard metrics, and (3) the shape of the retinal nerve fiber layer (RNFL) temporal, superior, nasal, inferior, temporal (TSNIT) curves obtained by the 2 different devices. METHODS: RNFL estimates were obtained from 136 eyes of 136 individuals (73 healthy and 63 mild glaucoma). WFA and FFA with and without asymmetry measures were performed on the TSNIT RNFL estimates to identify glaucoma from healthy eyes. Performance of WFA, FFA, and the standard metrics of OCT (Inferior Average) and GDX-VCC (Nerve Fiber Indicator) was evaluated by calculating receiver operating characteristic area. Measurements were obtained at a custom radius (33 to 41 pixels) for GDx-VCC to match the OCT radius (1.73 mm). RESULTS: WFA and FFA shape analysis significantly improved performance of both OCT (0.937) and GDx-VCC (0.913) compared with Inferior Average and Nerve Fiber Indicator (0.852 and 0.833, respectively). With either shape-based or standard metrics, OCT performance was slightly, but not significantly, better than GDx-VCC performance. Comparison of RNFL curves revealed that the GDx-VCC curves were more jagged and the peaks shifted more nasally when compared with the OCT RNFL curves. CONCLUSIONS: Performance of both OCT and GDx-VCC devices are improved by shape-based analysis methods. Classification performance was greater when using WFA for the OCT, and greater with FFA for the GDx-VCC. Significant differences between the machines exist in the measured TSNIT thicknesses, possibly because of GDx-VCC's measurements being affected by polarization magnitude varying with angle.