Retina blood vessel segmentation with a convolution neural network (U-net) takes up binary classification task to predicts if each pixel in the fundus image is either a vessel or not. The neural network architecture is derived from the U-net architecture, which achieves very good performance on very different biomedical segmentation applications. The neural network is trained on sub-images (patches) of the pre-processed full images. The loss function is the cross-entropy and the stochastic gradient descent is employed for optimization. It achieves the best score for DRIVE dataset in terms of area under the ROC curve in comparison to the other methods.
Input variables : Retinal Images
Output Variables : Pixel in the fundus image is a blood vessel or not
Visit Model : github.com
Additional links : arxiv.org
| Model Category | : | Public |
| Date Published | : | June, 2018 |
| Healthcare Domain | : | Medical Technology |
| Code | : | github.com |
| Medical Imaging |
| Image Synthesis |