advertisement
See also comment(s) by Kouros Nouri-Mahdavi •
PURPOSE: Measuring the spatial extent of defects may be advantageous in advanced glaucoma where conventional perimetric sensitivity measurements are unreliable. We test whether suprathreshold tests on a higher density grid can more efficiently map advanced visual field loss. METHODS: Data from 97 patients with mean deviation < -10 dB were used in simulations comparing two suprathreshold procedures (on a high-density 1.5° grid) to interpolated Full Threshold 24-2. Spatial binary search (SpaBS) presented 20-dB stimuli at locations bisecting seen/unseen points until the seen status of all neighbors matched or until tested points were adjacent. The SupraThreshold Adaptive Mapping Procedure (STAMP) presented 20-dB stimuli where entropy was maximal and modified the status of all points after each presentation, stopping after a fixed number of presentations (estimated as 50%-100% of the presentation number of a current procedure). RESULTS: With typical response errors, SpaBS had worse mean accuracy and repeatability than Full Threshold (both P < 0.0001). Compared to Full Threshold, mean accuracy (Full Threshold: median, 91%; interquartile range [IQR], 87%-94%) was slightly better with STAMP for all stopping criteria, although this was not statistically significant until 100% of conventional test presentations were used. Mean repeatability for STAMP was similar for all stopping criteria (P ≥ 0.02) compared to Full Threshold (Full Threshold: median, 89%; IQR, 82%-93%). CONCLUSIONS: STAMP accurately and repeatably maps the spatial extent of advanced visual field defects in as few as 50% of conventional perimetric test presentations. Further work is needed to test STAMP in human observers and in progressive loss. TRANSLATIONAL RELEVANCE: New perimetric approaches may improve information available for advanced glaucoma management and may potentially be more acceptable to patients.
School of Optometry and Vision Science, University of Bradford, Bradford, UK.
Full article