Abstract
Purpose :
Patients with ocular inflammatory disease often develop slowly progressive visual field loss. Visual fields, however, have inter-test variability and can be affected by media opacities and refractive errors. This study tested whether our automated OCT- Predictive Visual Threshold Sensitivity (OCT-PVST) can predict field loss compared to the Humphrey 24-2 (HVF 24-2).
Methods :
13 patients with idiopathic ocular inflammatory disease, confirmed by an uveitis expert and with known visual field defects on HVF 24-2 underwent pre-dilation Standard Humphrey 24-2 SITA perimetry. After dilation, our standard 9-field Spectralis OCT protocol was used, sequentially fixating on areas of retina 12.5° apart using a 3 × 3 grid pattern. This protocol entirely covers the 54° area tested with HVF. Using our automated pre-trained machine learning OCT-PVTS algorithm, which achieves 0.74 correlation with HVF 24-2 in patients with glaucoma, each 9-field OCT was co-registered and the nerve fiber, ganglion cell, and inner plexiform layers were co-segmented and each of the 52 HVF 24-2 testpoint thresholds was predicted and compared to HVF 24-2 measured thresholds.
Results :
Average correlation between actual and predicted HVF 24-2 thresholds was 0.53. As demonstrated in figure 1, predicted visual field corresponded reasonably well in most subjects. In several cases, the actual HVF 24-2 was abnormal while the prediction was relatively normal (i.e. the neuroretina had normal thickness), although the outer retina did appear thinner (as demonstrated in figure 2).
Conclusions :
Automated prediction of HVF 24-2 from 9-field OCT may have a role in management of patients with ocular inflammatory disease, if combined with analysis of the outer retina.
This is an abstract that was submitted for the 2017 ARVO Annual Meeting, held in Baltimore, MD, May 7-11, 2017.