Volumetric choroidal biomarkers in central serous chorioretinopathy using swept-source optical coherence tomography: a deep learning approach.

BACKGROUND

To investigate volumetric choroidal biomarkers in eyes with chronic central serous chorioretinopathy (cCSC) and healthy age-matched controls using a novel deep-learning approach.

METHODS

In this retrospective study, we included swept-source optical coherence tomography (SS-OCT) volumes of eyes affected by cCSC and healthy individuals. We obtained sector-wise choroidal volume biomarkers including choroidal vascularity index (CVI), choroidal vessel volume (CVV), choroidal volume (CV), and choroidal thickness (CT). Choroid layer segmentation was performed based on residual UNet deep learning method followed by volumetric smoothing method. Choroidal vasculature was obtained based on Phansalkar-thresholding-based method. T-test was used for the statistical analysis.

RESULTS

A total of 48 eyes (24 healthy and 24 cCSC) were included. No differences in age and sex were observed between the two groups (p > 0.05). CVV, CV and CT were significantly increased in cCSC eyes compared to controls in each sector (p < 0.05). On the other hand, CVI was significantly increased in the nasal and central sector in cCSC eyes compared to healthy age-matched individuals (0.33 ± 0.03 vs. 0.31 ± 0.04, p = 0.021; 0.33 ± 0.03 vs. 0.31 ± 0.03, p = 0.009 respectively), whereas no significant differences in the temporal, superior and inferior regions were observed.

CONCLUSION

cCSC eyes showed an increase in volumetric CT, CVV and CV in CSC eyes compared to healthy controls. Volumetric CVI was increased in the center and in the nasal sector in cCSC eyes compared to controls. Choroidal vascular volumetric analysis obtained with this novel deep learning approach will be further explored in other chorioretinal diseases.