An investigation into deep learning for locating the optic disc in scanning laser ophthalmoscope retinal images

The optic disc (OD) is a key anatomical structure in the and plays a fundamental role as a landmark in automatic screening systems. We present a deep learning approach for OD detection and segmentation in scanning laser ophthalmoscope (SLO) images. To train the model we exploit a transfer learning technique. The method reaches a mean Dice index of 0.91 with medical annotations. We analyse the medical consensus in indicating the OD and compare our method with other proposed in the literature.