Active Learning of Contrast Sensitivity Function to Assess Visual Outcomes in Age-related macular degeneration

Abstract

Purpose : To evaluate the application of active learning to measure the contrast sensitivity function in age-related macular degeneration (AMD).

Methods : Prospective, observational study performed at Mass Eye and Ear. We included eyes with dry AMD and wet AMD and excluded other visually significant diseases or previous ocular surgeries (except anti-VEGF intravitreal injection and cataract surgery). Eyes were tested with quick contrast sensitivity function (qCSF) using the Manifold platform (Adaptive Sensory Technologies, San Diego, CA) and spectral domain optical coherence tomography (Heidelberg). The main outcome measure was the area under log contrast sensitivity function (AULCSF). Secondary outcomes were contrast sensitivity thresholds at six spatial frequencies (1, 1.5, 3, 6, 12, 18CPD), contrast acuity (CA) and best corrected visual acuity (BCVA, LogMAR). All measures were compared to previously collected data in control eyes. After adjusting for sex and age, general linear models were used to compare the means of continuous variables.

Results : We included 40 eyes from 30 AMD patients and 30 eyes from 30 controls, mean aged 71.4±8.4 and 65.1±5.9 years old respectively. Sixteen eyes presented dry AMD, while 24 had wet AMD. Among the last, 13 eyes had fluid under the fovea. AMD eyes differed significantly from control eyes in BCVA (LogMAR, 0.13 VS. 0.01, P=0.001), mean AULCSF (0.75 ± 0.34 VS. 1.17 ± 0.26, P<0.001) and CA (1.04 ± 0.23 VS. 1.22 ± 0.13, P=0.023), after adjusting for sex and age. Eyes with dry AMD had a statistically significant reduction in AULCSF (P<0.05) despite no difference in visual acuity when compared to controls (P>0.05). There was also a non-significant reduction in AULCSF and CA in eyes with wet AMD compared to dry AMD (P>0.05). However, when looking at intermediate spatial frequencies, eyes with wet AMD had significantly reduced contrast thresholds compared to dry AMD at 1.5CPD, 3CPD and 6CPD (P<0.05).We found no statistically significant differences (P>0.05) in BCVA, mean AULCSF and CA between wet AMD patients with and without fluid under fovea.

Conclusions : An active learning algorithm reveals patterns of contrast sensitivity loss in eyes with AMD, which can be correlated with structural changes in AMD. These contrast sensitivity outcomes exhibit potential as endpoints in clinical trials for the treatment of AMD.

This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.