Abstract
Purpose :
Recent studies have promoted EZ area as an effective, reliable surrogate biomarker for visual function in retinal degenerations. Reliability studies may be too optimistic if area alone is considered because dissimilar shapes can have similar areas. Here, we analyze the utility of two EZ area measurement methods and the influence of EZ shape.
Methods :
EZ boundaries in volumetric SDOCT scans were marked in 88 eyes from 50 patients in the Trial of Oral VPA for Retinitis Pigmentosa (NCT01233609) by three trained graders (2 expert, 1 novice) using two methods: transverse and enface. For transverse, EZ endpoints were marked on each b-scan (similar to Ramachandran et al, TVST, 2013 and 2016). For enface, EZ boundaries were traced on enface maximum intensity projection images (similar to Hariri et al, JAMA Ophthalmol, 2016). Each method produced an EZ boundary polygon and associated EZ area (Fig 1a,b). EZ area differences (between graders, over time, etc) were compared by simple subtraction of areas, and by a shape-sensitive approach using the area of symmetric difference (ASD). The ASD of two polygons is the area of their union minus their intersection, and detects differences in differently shaped EZ regions even if they have similar areas (Fig 1c).
Results :
Table 1 reports grading times and reliability metrics. The transverse method was the most reliable but only among expert graders, while the enface method was about twice as fast with better agreement among all graders. After accounting for shape differences with ASD, grader agreement for both methods was still good to excellent, but significantly reduced (p<0.001 for all comparisons) and, for the enface method, much lower than published studies based on area alone. Fig 1d shows the percent change in EZ boundary over one year plotted by retinal direction, an example result made possible by shape analysis.
Conclusions :
EZ boundary shape should be considered when quantifying differences and longitudinal changes in EZ area. Doing so will resolve the geometric ambiguity of the area, improve change detection sensitivity, and provide new opportunities for EZ change analysis.
This is an abstract that was submitted for the 2017 ARVO Annual Meeting, held in Baltimore, MD, May 7-11, 2017.