DRUSEN MEASUREMENTS COMPARISON BY FUNDUS PHOTOGRAPH MANUAL DELINEATION VERSUS OPTICAL COHERENCE TOMOGRAPHY RETINAL PIGMENT EPITHELIAL SEGMENTATION AUTOMATED ANALYSIS

Abstract

To compare drusen measurements obtained from color fundus and infrared retromode photographs with those derived from spectral-domain optical coherence tomography. Drusen lesions identified on the planar (color and infrared) imaging modalities were manually segmented by two independent graders using previously described reading center software to produce quantitative measurements of drusen area and number. The corresponding volume Cirrus OCT datasets were analyzed using commercial retinal pigment epithelium analysis algorithms to segment the retinal pigment epithelium band and estimated the drusen area. Drusen numbers were extracted from retinal pigment epithelium elevation maps. Intraclass correlation coefficients assessed agreement between graders; graders' average measurements were compared with optical coherence tomography (OCT) using paired T-tests. Excellent agreement between graders was observed (r = 0.951-0.974). No statistical difference was found in the area values obtained by color (0.85 ± 0.26 mm(2), P = 0.43) or retromode (1.15 ± 0.32 mm(2), P = 0.35) compared with those obtained by OCT (0.98 ± 0.28 mm). The number of drusen identified by OCT (13.15 ± 3.19) was significantly lower than that determined by manual segmentation of color (53.7 ± 13.18) and retromode (100.13 ± 16.18) images. Although the number of drusen individualized by commercial OCT algorithms is significantly lower than by planar fundus imaging modalities, the OCT-measured drusen area is not affected, suggesting that the algorithm counts confluent drusen as a single drusen.