Sensitivity and Specificity of Reticular Pseudodrusen detection using multimodal imaging in Amish Population

Abstract

Purpose : To compare sensitivity and specificity for the detection of reticular pseudodrusen (RPD) associated with age-related macular degeneration using multiple imaging modalities.

<gdiv></gdiv><gdiv></gdiv>

Methods : A post-hoc analysis was performed on images collected from a family-based prospective cohort study of 1320 elderly Amish subjects (age range 50–99 years) (2,640 eyes) who had a family history of at least 1 individual with AMD. All subjects underwent complete ophthalmic examination, spectral domain optical coherence tomography (SDOCT; 6x6mm ~ 20^o), blue-light fundus autofluorescence (FAF), infrared reflectance (IR), and flash color fundus photography (CFP). Both eyes were included in this analysis. Individual imaging modalities were assessed separately in a masked fashion by expert human graders for the presence of RPD, also termed subretinal drusenoid deposits (SDD). To be deemed to be present, a minimum of three discrete lesions were required. Ground truth was established based on the presence of RPD on at least two modalities. Sensitivity and specificity were computed for each individual modality against the ground truth. As SDOCT images covered a smaller region (20^o vs 30^o) compared to other imaging analyses, SDOCT sensitivity and specificity was recomputed after cropping the other modalities to a similar region of interest.

<gdiv></gdiv><gdiv></gdiv>

Results : RPD were noted to be present on at least one modality in 140 (6.3%) of the 2640 eyes, and in 132 eyes (of 66 subjects; 5%) they were noted in at least two modalities (ground truth or reference determination). Overall, RPD were observed in 133 (5 %), 126 (4.8%), 115 (4.4%), 35 (1.3 %) eyes by IR, FAF, SDOCT and CFP respectively. Sensitivity and specificity for identification of RPD for each imaging modality is shown in Table 1. IR demonstrated the highest sensitivity, and CFP showed the lowest sensitivity. When cropping to a similar region, sensitivity and specificity of SDOCT improved to 93% and 95%, respectively.




<gdiv></gdiv><gdiv></gdiv>

Conclusions : IR imaging appears to be the most sensitive modality for detection of RPD, with a high specificity. SDOCT imaging can also show high sensitivity but may require the scan field to be enlarged beyond the dimensions commonly acquired in clinical practice.
<gdiv></gdiv><gdiv></gdiv>

This is a 2021 ARVO Annual Meeting abstract.