Automated segmentation of retinal neuroepithelial layer and pigment epithelium in macular OCT images using U-Net and ISC

Retinal optical coherence tomography (OCT) images can be used for segmenting the retinal neuroepithelial layer and pigment epithelium. Segmentation is important for both ophthalmic diagnosis and the study of eye diseases. We propose an automated segmentation method for the retinal neuroepithelial layer and pigment epithelium in macular OCT images based on the widely used U-Net machine learning architecture and recently proposed integration of self-attention and convolution (ISC) approach. To verify the effectiveness of our method, we performed an experiment and compared its results with the corresponding ground truths in qualitative and quantitative terms. Our method is able to accurately segment the retinal neuroepithelial layer and pigment epithelium in macular OCT images, as demonstrated by the experimental results.