Abstract
Purpose: :
The outer boundary line (BL) location differs among SDOCT vendors making it difficult to compare retinal thickness data in clinical trials. No automated software is commercially available that can segment a common BL from all SDOCT types. Manual segmentation of BLs is effort intensive using vendor’s software. We evaluated the accuracy and efficiency of caliper measurement to generate a central subfield (CSF) thickness value comparable to the CSF generated from manual segmentation.
Methods: :
SDOCT raster volume scans from 41 eyes with various retinal diseases were taken with Spectralis or Cirrus machines were evaluated in the vendor’s software. The BLs were redrawn to top of RPE in all B scans in the CSF. Caliper measurement was performed (ILM -RPE) at the center point (CPT) and 500µ from the center in 4 directions vertically and horizontally. CSF was generated in 3 different ways: 1) A regression equation; imputed CSF = 0.85xCPT+58.6 (similar to DRCR network approach), imputes the CSF from a single CPT measurement. This equation was developed on an independent SDOCT dataset using the correlation between machine generated CPT and CSF thickness. 2) The 5 caliper measured points were averaged to calculate CSF. 3) Machine regenerated CSF after manual segmentation, the "gold standard".
Results: :
The mean imputed CSF was 371µ (199), 5 point CSF was 392 µ (192) and manually segmented CSF was 390µ (197). The mean difference between imputed CSF and manually segmented CSF was 19.55 µ (limits of agreement, -34.66, 73.76). The mean difference between 5 point CSF and manually segmented CSF was 2. 24 µ (limits of agreement, -30.96, 26.48). Imputed CSF differed by >20 µ from manual segmentation in 52%, whereas 5 point CSF differed by >20 µ from manual segmentation in only 12% of scans.
Conclusions: :
The difference between 5 point CSF and manually segmented CSF is well within 20µ, the reported repeatability in SDOCT measurements. Until a solution towards a common outer BL can be found for SDOCTs, the 5 point caliper measurement is a potential alternative solution that should be evaluated in a larger dataset.
Keywords: imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • retina • clinical research methodology