Abstract
Purpose :
The newest generation of Optical Coherence Tomography (OCT) devices promise faster acquisition times and improved image quality. However, switching instruments can cause discontinuities in longitudinal series. In this study, we evaluate the effect of upgrading the Spectralis OCT (Heidelberg Engineering, Germany) to the 85kHz OCT2 module.
Methods :
27 participants in the Portland Progression Project were scanned using the OCT1 (software version 5.6.1) and using the new OCT2 module (software version 6.8.3), in both eyes, on the same day. All participants had suspected or evident glaucoma as judged by their clinician. Peripapillary circle scans at 3.5mm radius, and 24 radial scans cerntered on the optic nerve head, were performed on both insturments. The manufacturer's software was used to generate retinal nerve fiber layer thickness (RNFLT) and minimum rim width (MRW), using default settings, globally and for each of the software's six predefined sectors. These were compared using Bland-Altman plots. The difference between instruments was compared against within-eye inter-visit limits of agreement for OCT1, based on a separate cohort of 54 eyes for RNFLT and 16 eyes for MRW tested five times within a few weeks.
Results :
The mean global average RNFLT was 77μm for OCT1, and 79μm for OCT2. Mean MRW was 216μm for OCT1 and 219μm for OCT2. The top part of each figure shows a Bland-Altman plot comparing global average thicknesses between instruments, together with the line of best fit (in red) plus/minus the within-eye limits of agreement for the OCT1 (in blue). Differences were greater in healthier eyes. Even accounting for this bias, RNFLT and MRW from OCT2 were outside the OCT1 limits of agreement in 56% and 26% of eyes respectively. Segmentations were manually-corrected for scans used to derive those limits, but not for the scans used in the comparison. The two outliers for MRW were both caused by inaccurate automated segmentations for the OCT1 scan; both were within the blue lines after manual correction. The largest difference between instruments in global RNFLT was 12μm, where 3.84μm is one pixel; but differences were >20μm in 5.2% of sectors.The bottom part of each figure shows how much thicker each sector would appear to be using OCT2, for an eye with median thickness by OCT1.
Conclusions :
OCT2 tended to give higher values for both RNFLT and MRW than OCT1, especially in healthier eyes.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.