Abstract
Purpose :
To estimate sensitivity, specificity, positive and negative predictive value of optical coherence tomography angiography (OCTA), fluorescein angiography (FA) and indocyanine green angiography (ICGA) in detecting choroidal neovascularization (CNV) complicating geographic atrophy (GA).
Methods :
Consecutive patients affected by age related macular degeneration with GA and GA complicated by CNV were collected by a senior reader with a multimodal imaging approach including FA, ICGA, SD-OCT and OCTA. Detection rates of CNV were then evaluated by two masked independent readers using FA, ICGA and OCTA, separately. In particular, each image was shuffled and presented to the readers in random order. OCTA was assessed using manual and automated segmentation, including outer retina to choriocapillaris (ORCC), avascular, choriocapillaris and choroid.
Results :
Forty-two patients were enrolled, divided in 21 with GA and 21 with GA complicated by CNV. In the group of patients with CNV and GA, 77% presented an active CNV, while 23% presented no signs of activity. No differences were observed between the two groups for age, sex and best corrected visual acuity (p> 0.05).
Sensibility, specificity, positive and negative predictive value for each imaging modalities are presented in Table 1. The highest values were recorded by manual and ORCC segmentations of OCTA. Cohen's kappa coefficient was used to evaluate agreement between the readers (0.84 for FA, 0.87 for ICGA and 0.93 for OCTA).
Conclusions :
Detecting CNV in GA with dye based imaging is often challenging due to the low contrast between CNV and the underneath choroid. On the contrary, OCTAis not influenced by dye leakage and allows the identification of different retinal layers separately. OCTA demonstrated the highest sensibility, specificity, positive and negative predictive values compared to the other imaging modalities. For this reason, OCTA should be used in the multimodal imaging evaluation to asses the presence of CNV in GA. In this contest, it is essential to manually adjust the segmentation boundaries to detect the lesion.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.