Abstract
Purpose:
The objective of this project is to develop, test and validate a software allowing to estimate photoreceptor density derived from the inner segment ellipsoid (ISe) band reflectivity on en-face OCT.
Methods:
Eight eyes displaying various degrees of outer retinal impairment (five eyes with resolved central serous chorioretinopathy, one with hydroxychloroquine toxicity, and two healthy eyes) were imaged using a spectral-domain OCT (Spectralis, Heidelberg, Heidelberg Engineering, Germany). En-face OCT was generated from horizontal transverse scans, 11µ apart, with a resolution of 384A scan/Bscan. Segmentation of the Ise band was automatically performed with the Heidelberg Explorer plug-in provided by the manufacturer. The relevance of the “PR1” segmentation line was checked by two readers. At the same time, a software running on Matlab, (Matlab, Mathworks Inc., Natick, USA) that transforms the grey level into cone density was developed. Its algorithm is based on a previous work comparing OCT reflectivity and adaptive optics. The program determines the cone density in a ring ranging from 1.5 to 2 degrees of eccentricity from the fovea. In order to validate the estimates, cone densities were measured in the same eyes, in a blind manner, by an independent observer, using adaptive optics (AOdetect v0.2, Imagine Eyes, Orsay, France).
Results:
Thirty five points of comparison were obtained. The program estimates were closely correlated with the results of the reference counting method, i.e. adaptive optics (r2=0.81, p<0.05). Bland-Altman plot displayed a moderate estimated bias of 204±2660 cellules/mm2(CI95%: 5008-5418). The estimatation was considered relevant (defined as a difference of more and less 10%) in 80% of the cases.
Conclusions:
It has become possible to estimate cone density with an OCT device currently used in routine practice, at least in a defined retinal area. Improvement in the acquisition procedure (resolution, artefact management...) will eventually improve the accuracy of the results.