Abstract
Purpose :
Testing central visual fields (VF) densely with a 10-2 test pattern can locate glaucomatous damage that a sparser grid may miss. However, it is often not feasible to do both a 10-2 and 24-2 test, nor is it obvious when to switch from one to another for a given patient. Here we explore whether the published ARREST approach to VF testing (Turpin et al., TVST, 2018) can be used to automatically incorporate 10-2 locations into the 24-2 without increasing test times nor harming the ability to detect glaucomatous progression.
Methods :
ARREST uses the previous VF test to choose locations to test in the current test, censoring locations < 17dB (which are screened to ensure not < 0dB), adding new locations in regions of interest (ROI) without increasing test times. Here, we set the ROI to 10-2 locations located on the steepest gradient of the previously tested VF, creating ARREST-10. New locations were added when at least one location in the central 9 degrees fell below 17dB. We compare sensitivity of ARREST-10 for detecting progression with that of a 24-2 pattern. We simulate both approaches on 12 visits of 107 eyes whose final field is known, with prior fields back calculated using the method in Turpin et al., TVST, 2018. This creates a synthetic dataset where 100% of the VF series are progressing. Progression is determined using the PoPLR method, with event-based criteria ignored in ARREST-10. Specificity is checked by simulation on the final VF repeated 5 times.
Results :
Final fields had median Mean Total Deviation of -6.0 dB (95% range (-16.9, 0.2)). ARREST-10 used on average 23 less presentations on the final field than ZEST (paired t.test p < 0.001). ZEST had slightly higher sensitivity at visits 4 and 5 for matched specificities (Figure 1). 57 of the 107 eyes had no extra locations added to the 24-2 by ARREST-10, with the remainder having between 1 and 30 10-2 locations added by the 12th visit. Figure 2 shows the final locations tested for one eye.
Conclusions :
Using the 10-2 as the ROI in the ARREST approach allows automatic incorporation of 10-2 locations into the 24-2 pattern customised to individuals without increasing test times.
This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.