Signal Normalization Reduced Systematic Differences in Retinal Nerve Fiber Layer (RNFL) Thickness Measurements Between Spectral Domain Optical Coherence Tomography (SD-OCT) Devices

Abstract

Purpose: : Differences in RNFL thickness measurements between SD-OCT devices pose significant difficulty in both clinical and research settings where multiple devices are in use. The purpose of this study was to develop and test a novel signal normalization method that reduced systematic differences.

Methods: : Ninety-six eyes (32 glaucomatous, 33 glaucoma suspect, and 31 healthy eyes) from 51 subjects were scanned with two SD-OCT devices on the same day (Cirrus HD-OCT; Carl Zeiss Meditec, Inc., Dublin, CA; Optic Disc Cube 200x200 scan pattern, and RTVue; Optovue, Fremont, CA; RNFL 3.45 Circle scan pattern). RTVue data had 768 pixels for 2.3 mm scan length on each A-scan, which was oversampled and padded with minimum signal to match with the Cirrus specification (1024 pixels for 2.0 mm). Then both the original Cirrus signal and RTVue derived Cirrus signal were normalized in amplitude so that meaningful signal range was stretched to the full gray scale range of 0 to 255. To compensate signal strength difference, custom high dynamic range (HDR) processing was also applied to poor signal strength images (lowest 20 percentile of the entire set of images). Conventional global mean circumpapillary RNFL thicknesses were measured automatically using software of our own design and then compared to the original device outputs. Measurement error models were used to analyze the absolute RNFL thickness difference between original device outputs and our software outputs after signal normalization. P<0.05 was considered as statistically significant.

Results: : The mean absolute difference in RNFL thicknesses between Cirrus and RTVue were significantly different before normalization (10.5 ± 4.1 μm), but not after normalization (6.1 ± 4.7 μm). The absolute difference of RNFL thickness between Cirrus and RTVue was statistically significantly reduced by signal normalization. The median of the absolute difference of the RNFL thickness after normalization was 4.7 μm, which is within the inherent device measurement variability.

Conclusions: : The reported novel signal normalization method successfully reduced the systematic difference in RNFL thickness measurement between Cirrus and RTVue to the level of the inherent device measurement variability. Enabling direct comparison of RNFL thickness obtained from multiple devices would broaden the use of OCT technology in both clinical and research applications.

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