Purchase this article with an account.
Leon von der Emde, Maximilian Pfau, Sarah Thiele, Ruth Hassenrik, Philipp T Möller, Jennifer Nadal, Matthias Schmid, Monika Fleckenstein, Frank Holz, Steffen Schmitz-Valckenberg; Structure-function correlation of dark-adapted two color fundus-controlled perimetry and multimodal imaging in neovascular age-related macular degeneration. Invest. Ophthalmol. Vis. Sci. 2018;59(9):4945.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To investigate the functional impact of structural changes on mesopic and scotopic retinal sensitivity in neovascular age-related macular degeneration (nAMD).
Twenty eyes of 20 nAMD patients (mean age 76.8 years) and 40 eyes of 40 normal subjects underwent multimodal imaging and duplicate mesopic and two-color dark-adapted (DA) perimetry using the S-MAIA device (S-MAIA, CenterVue, Italy). The outer nuclear layer (ONL), inner (IS) and outer photoreceptor segments (OS) and retinal-pigment-epithelium-drusen-complex were semi-automatically annotated in spectral domain optical coherence tomography (SD-OCT) scans (121 B-scans, 30°x25°, Spectralis OCT2, Heidelberg Engineering, Germany). FCP data were registered to SD-OCT data to obtain thickness data spatially corresponding to the stimulus-location and area (0.43°). Using random forest (RF) regression, we evaluated (i) the percentage of variability in retinal sensitivity explained by the models with the R2 criterion with and without patient-specific training data (leave-one-out cross validation on patient level [LOO-CV]) and (ii) the relative importance of layers with the average increase in out-of-bag mean-squared error (%IncMSE).
In eyes with nAMD, the mean deviation (mean ± SD) was -7.7 ± 5.5 dB for mesopic, -9.8 ± 5.0dB for DA cyan and -7.9 ± 4.9dB for DA red testing. For mesopic testing, RF regression could explain without LOO-CV 50.5 % and with LOO-CV 24.1 % of the variability in point-wise sensitivity. Hereby, the ONL (19.1 %IncMSE) and OS thicknesses (16.9 %IncMSE) were the most important predictive features. For DA cyan and red testing, the R2 values were without LOO-CV 40.8 % and 51.5 % or with LOO-CV 22.5 % and 30.6 %. Again, the ONL (18.3 and 14.5 %IncMSE) and OS thicknesses (13.7 and 15.6 %IncMSE) were the most important predictive features.
The larger deviation for DA cyan testing as compared to DA red testing is indicative of a more severely impaired rod than cone function in nAMD. Volumetric SD-OCT data may explain a significant proportion of the variability of mesopic and scotopic dysfunction. Besides layer thicknesses, SD-OCT reflectivity profiles and fundus autofluorescence data may provide even more accurate prediction of retinal sensitivity impairment.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.
This PDF is available to Subscribers Only