We have developed and trained a convolutional neural network to automaticallyand simultaneously segment optic disc, fovea and blood vessels. Fundus imageswere normalised before segmentation was performed to enforce consistency inbackground lighting and contrast. For every effective point in the fundusimage, our algorithm extracted three channels of input from the neighbourhoodof the point and forward the response across the 7 layer network. In average,our segmentation achieved an accuracy of 92.68 percent on the testing set fromDrive database.
展开▼
