Abstract
Purpose: :
To evaluate the specific automated detection and quantification of hard exudates (HE) in diabetic macular edema (DME) in vivo using polarization sensitive OCT (PS-OCT).
Methods: :
In this pilot study, 10 eyes with DME were examined by color fundus (CF) photography and by a PS-OCT device. The PS-OCT, currently being in a prototype state, allows raster scanning of the retina enabling to detect, map and measure retinal structures based on their polarization scrambling qualities. Each eye was examined 5 times in a row, providing comparable data in respect of image quality, position and state of the DME. From these data, 2D and 3D thickness maps of the respective HE were generated and then correlated with the according CF photographs.
Results: :
While correlation of the two different image types (CF and 2D mapping by PS-OCT) was strong with a Pearson's r = 0.9 (p < 0.001), agreement by Bland and Altman was modest. By PS-OCT, HE showed an increased detection-rate with a mean of 31.64% (95% CI 14.29% - 48.99%) compared to CF imaging. Analysis of reproducibility of the HE quantification by PS-OCT yielded solid results with a coefficient of variance ranging from 0.19 to 0.36 and an ICC of 0.885 (95% CI 0.665 - 0.985).
Conclusions: :
This study presents a new and reliable way to automatically detect HE due to their polarization scrambling properties in PS-OCT. Further, HE can now be quantified not only by the area they cover in the fundus two-dimensionally but by their full extent using 3D-mapping. Moreover, the increased detection of HE by PS-OCT suggests a higher sensibility compared to CF photography. This is lead back to PS-OCT device's quality to detect already precursors of HE before being detectable by conventional CF imaging. By using PC-OCT for the quantification of HE, new possibilities are offered for analysis of DME morphology and clinical follow-up.
Keywords: diabetic retinopathy • imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • imaging/image analysis: clinical