Self-supervised retinal thickness prediction enables deep learning from unlabelled data to boost classification of diabetic retinopathy.
Olle G. HolmbergNiklas D. KöhlerThiago MartinsJakob SiedleckiTina HeroldLeonie KeidelBen AsaniJohannes SchiefelbeinSiegfried PriglingerKarsten U. KortuemFabian J. TheisPublished in: Nat. Mach. Intell. (2020)
Keyphrases
- deep learning
- unlabelled data
- labelled data
- unsupervised learning
- retinal images
- supervised and unsupervised learning
- diabetic retinopathy
- machine learning
- co training
- supervised learning
- pattern recognition
- training set
- decision trees
- learning phase
- feature vectors
- support vector machine
- text classification
- data sets
- computer vision
- clustering algorithm
- feature extraction
- early detection
- weakly supervised
- blood vessels
- classification algorithm
- model selection
- multi view
- active learning