Automatic 3D bi-ventricular segmentation of cardiac images by a shape-constrained multi-task deep learning approach.
Jinming DuanGhalib BelloJo SchlemperWenjia BaiTimothy J. W. DawesCarlo BiffiAntonio de MarvaoGeorgia DoumouDeclan P. O'ReganDaniel RueckertPublished in: CoRR (2018)
Keyphrases
- deep learning
- multi task
- shape prior
- learning tasks
- unsupervised learning
- machine learning
- level set
- segmentation algorithm
- shape model
- segmentation method
- medical images
- learning problems
- object segmentation
- multiscale
- mental models
- multi class
- active contours
- image segmentation
- higher order
- training data
- transfer learning
- probabilistic model
- feature extraction
- support vector machine
- information gain
- data mining
- feature selection
- superpixels
- image classification
- weakly supervised
- supervised learning