Sensitivity and Specificity of Multimodal Imaging in Characterizing Drusen

Abstract

To calculate the measures of accuracy of different imaging modalities in patients with early/intermediate age-related macular degeneration (AMD). Prospective, observational, cross-sectional study. Patients with early or intermediate AMD. All participants underwent a complete multimodal imaging assessment with a confocal scanning laser ophthalmoscope, including near-infrared reflectance (NIR), green fundus autofluorescence (G-FAF), confocal pseudocolor, and retromode deviated to right (DR) and left (DL). Drusen were topographically divided as small and medium (≤125 μ diameter) and large (>125 μ diameter), whereas subretinal drusenoid deposits (SDDs) were divided into dot and ribbon phenotypes. Multimodal imaging was considered the reference standard for detecting different subtypes of drusen and SDDs. Cohen's kappa (k) was used to test interobserver agreement for each imaging modality. Capability to differentiate subtypes of drusen and SDDs with different imaging modalities. A total of 100 eyes (62% were female participants) were evaluated. The inter-rate reliability between 2 readers for each imaging modality ranged between 0.76 and 0.95. Overall, large drusen were better identified with confocal pseudocolor imaging (96.6% sensitivity; 77.8% specificity). Smaller drusen were better detected with retromode modalities DR or DL (92% sensitivity; 58.3% specificity and 85.2% sensitivity; 83.3% specificity, respectively). Ribbon SDDs were better detected by color imaging (80.5% sensitivity; 98.3% specificity). Dot SDDs were well identified with NIR (83.1% sensitivity; 91.4% specificity) and G-FAF (84.6% sensitivity; 77.1% specificity). Near-infrared reflectance and G-FAF should be considered for classification of dot SDDs, and confocal pseudocolor is optimal for characterizing ribbon SDDs. Among all imaging modalities, retromode technology DR and DL may be a potential supplementary modality to detect even smaller drusen.