Deep learning-based 3D OCT imaging for detection of lamina cribrosa defects in eyes with high myopia

Abstract

The lamina cribrosa (LC) is a collagenous tissue located in the optic nerve head, and its dissection is observed in eyes with pathologic myopia as a LC defect (LCD). The diagnosis of LCD has been difficult because the LC is located deep beneath the retinal nerve fibers. The purpose of this study was to determine the prevalence and three-dimensional shape of LCDs in highly myopic eyes. Swept-source OCT scan images of a 3 × 3-mm cube centered on the optic disc were obtained from 119 eyes of 62 highly myopic patients. Each LC was manually labelled in cross-sectional OCT images along the axial, coronal, and sagittal planes. A deep convolutional neural network (DCNN) was trained with the manually labeled images, and the trained DCNN was applied to the detection of the LC in every image in each plane. Three-dimensional images of the LC were generated from the labeled image of each eye. The results showed that LCDs were detected in 12 of the 42 (29%) eyes in which an LC was visible. The LCDs ran vertically at the temporal edge of the optic disc. In conclusion, 3D OCT imaging with the application of DCNN is helpful in diagnosing LCDs.