Retinal Thickness Measurements with Spectral Domain Optical Coherence Devices from Different Manufacturers in a Reading Center Environment

Abstract

Purpose: : Central retinal thickness (CRT) measured by optical coherence tomography (OCT) is used as an endpoint variable in many multicenter trials on retinal diseases. As different clinical sites use OCT equipment from different manufacturers a standardized evaluation of OCT scans from different devices is of great importance. Therefore the Vienna Reading Center (VRC) has developed custom software to enable evaluations of OCT scans from several manufacturers within one application. A validation of this approach shall be performed by scanning patients with different retinal pathologies with OCT devices from selected manufacturers and comparing CRT values derived from each instrument. OCT devices not correlating with the other instruments shall be identified.

Methods: : 29 eyes of 19 patients (19 eyes with macular edema secondary to diabetic maculopathy, central/branch vein occlusion or age related macular degeneration, 6 eyes with other macular diseases and 4 healthy eyes) were imaged with four common used spectral domain OCT devices, the Zeiss Cirrus HD-OCT (CIRR), Heidelberg Spectralis OCT (SPEC), Topcon 3D-OCT 2000 (TOP2) and Nidek RS-3000 (NID3). Thickness of the central mm was evaluated by certified graders of the VRC for all four OCT devices from the original instrument (To), after manual correction of the center point (Tcp), after correction of center point and alignment errors (Tcpal) and after correction of center point and alignment errors with the VRC customized software (Tvrc). SPEC, TOP2 and NID3 were always compared to CIRR using the Concordance Correlation Coefficients (CCC) and Bland Altman-plots.

Results: : For CIR minimum, maximum and mean values in µm were 95/600/345 for To, 94/620/340 for Tcp, 96/620/343 for Tcpal and 108/658/383 for Tvrc. For To and Tvrc minimum/maximum/mean values were in µm 22/640/328 and 112/668/378 for SPEC, 211/563/329 and 104/635/367 for TOP2 versus 119/622/356 and 119/637/369 for NID3. Correlating CIRR to the other three instruments the Pearson (precision) value of the CCC ranged from 0.460 to 0.945 for To, from 0.977 to 0.986 for Tcp and from 0.983 to 0.989 for Tvrc. The bias correction factor (accuracy) of the CCC ranged from 0.977 to 0.995 for To, from 0.975 to 0.998 for Tcp and from 0.992 to 0.999 for Tvrc.

Conclusions: : Correlations of SPEC, TOP2 and NID3 with CIRR were good for all measurement methods but best when achieved with the VRC customized software including correction of the center point and correction of alignment errors. In scans showing retinal pathology manual correction of the center point can be challenging even for certified graders but is of great importance for accuracy of thickness values in multicenter trials.

Keywords: imaging/image analysis: clinical • imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • clinical research methodology