Thickne Profile of Retinal Layer by Optical Coherence Tomography Image egmentation Original Reearch Article

Purpoe To report an image egmentation algorithm that wa developed to provide quantitative thickne meaurement of ix retinal layer in optical coherence tomography (OCT) image. Deign Propective cro-ectional tudy. Method Imaging wa performed with time- and pectral-domain OCT intrument in 15 and 10 normal healthy ubject, repectively. A dedicated oftware algorithm wa developed for boundary detection baed on a 2-dimenional edge detection cheme, enhancing edge along the retinal depth while uppreing peckle noie. Automated boundary detection and quantitative thickne meaurement derived by the algorithm were compared with meaurement obtained from boundarie manually marked by three oberver. Thickne profile for ix retinal layer were generated in normal ubject. Reult The algorithm identified even boundarie and meaured thickne of ix retinal layer: nerve fiber layer, inner plexiform layer and ganglion cell layer, inner nuclear layer, outer plexiform layer, outer nuclear layer and photoreceptor inner egment (ONL+PI), and photoreceptor outer egment (PO). The root mean quared error between the manual and automatic boundary detection ranged between 4 and 9 μm. The mean abolute value of difference between automated and manual thickne meaurement were between 3 and 4 μm, and comparable to interoberver difference. Inner retinal thickne profile demontrated minimum thickne at the fovea, correponding to normal anatomy. The OPL and ONL+PI thickne profile repectively diplayed a minimum and maximum thickne at the fovea. The PO thickne profile wa relatively contant along the can through the fovea. Concluion The application of thi image egmentation technique i promiing for invetigating thickne change of retinal layer attributable to dieae progreion and therapeutic intervention.