Purchase this article with an account.
Siamak Yousefi, Hiroshi Murata, Yuri Fujino, Ryo Asaoka; An individualized test to detect glaucoma progression in visual fields. Invest. Ophthalmol. Vis. Sci. 2017;58(8):2865. doi: https://doi.org/.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To detect glaucomatous progression in a longitudinal series of visual fields (VFs) by combining the significance of progression in all VF test locations
VFs from 106 eyes of 71 participants in the University of Tokyo hospital were obtained using the Humphrey 30-2 or 24-2 VF tests. Eyes with at least 16 visits with approximately six-month interval between visits (interval greater than 3 months and less than 2 years) were included in the study. All 30-2 tests were matched to 24-2 tests resulting in 52 VF test locations for all eyes (excluding blind spots). Linear regression of each test location was computed and the significance of slopes (p-value) was calculated. All 52 p-values were combined using a modified version of Wilkinson approach (median of the significant points) to generate a single p-value. The accuracy of the progression detection was assessed on VF series of five visits to 15 visits (excluding the first visit) using the hit rate, that was computed on the original VF series, versus false positive rate (FPR), which was computed on the randomly permuted VF series. The proposed approach was compared against linear regression of mean deviation (MD) and permutation of point-wise linear regression (PoPLR) employing the receiver operating characteristic (ROC) curves.
The partial area under the ROC curve (from 0 to 0.15 FPR) was 0.02 for linear regression of MD, PoPLR, and the proposed approach using five visits while the partial area under the ROC curve was 0.08 for linear regression of MD and PoPLR and 0.10 for the proposed approach using 15 visits.
Accuracy of the linear regression of MD was comparable to accuracy of the PoPLR and the proposed approach using five visits. The accuracy of the proposed approach outperformed that of the linear regression of MD and PoPLR at higher number of visits.
This is an abstract that was submitted for the 2017 ARVO Annual Meeting, held in Baltimore, MD, May 7-11, 2017.
Figure 1. Glaucoma progression detection. Left panel shows progression detection of linear regression of MD, the middle panel shows the progression detection of PoPLR, and the right panel shows the progression detection of the proposed approach for VF series of five visits to 15 visits.
This PDF is available to Subscribers Only