Retinal vessel segmentation of color fundus images using multiscale convolutional neural network with an improved cross-entropy loss function.
Kai HuZhenzhen ZhangXiaorui NiuYuan ZhangChunhong CaoFen XiaoXieping GaoPublished in: Neurocomputing (2018)
Keyphrases
- loss function
- optic disc
- cross entropy
- fundus images
- multiscale
- segmentation scheme
- retinal vessel
- convolutional neural network
- blood vessels
- pairwise
- retinal images
- support vector
- image segmentation
- log likelihood
- learning to rank
- segmentation algorithm
- diabetic retinopathy
- maximum likelihood
- vessel segmentation
- segmentation method
- image processing
- low contrast
- superpixels
- edge detection
- optical coherence tomography
- face detection
- evaluation metrics
- level set
- image analysis
- vessel enhancement