Purchase this article with an account.
Chahira Miloudi, Laurent Mugnier, Jose Alain Sahel, Florence Rossant, Isabelle Bloch, Michel Paques; Adaptive optics photoreceptor mapping : Compensation of photoreceptor scintillation by image fusion. Invest. Ophthalmol. Vis. Sci. 2015;56(7 ):4940.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Adaptive optics (AO) images of photoreceptors are affected by spatial and temporal variability. In particular, high frequency (>10Hz) variation of reflectance of individual cones, termed here scintillation, may be observed. We hypothesized that scintillation is a major source of photoreceptor reflectance variability and hence of cone counting. In order to improve the reliability of cone countings, we developed a procedure for compensating scintillation based on fusion of deconvolved AO images.
AO fundus images were obtained through dilated pupils with a commercially available flood imaging AO camera (rtx1™ camera; Imagine Eyes, Orsay, France; illumination wavelength 840nm) within an IRB-approved clinical study in 5 normal eyes. The routine acquisition procedure comprised a stack of 40 raw images acquired over 4.2 seconds, 2° from the fovea. For each subject, raw images were deconvolved with a myopic deconvolution method [Blanco and Mugnier, 2011] and underwent automatic cone detection [Loquin et al. 2011] over a 94x94µm area. Cone maps obtained from raw images were fused; the increment in cone density at every step was measured. The results were compared to cone maps obtained by image averaging (AO image 2.0)
In three subjects, the increment in cone density reached an asymptotic plateau between the 10th and 20th map integration. In one subject the plateau was not obtained (i.e. the addition of new raw images increased the total amount of cone detected). The last subject had limited scintillation, hence there was limited increment. The relative improvement in total number of detected cone by fusing 20 raw deconvolved images compared to the number of cones detected on averaged image ranged from 5% to 71%.
Fusion of deconvoluted raw images improves the quality of cone detection from flood-illuminated AO images.
This PDF is available to Subscribers Only