MetricUNet: Synergistic image- and voxel-level learning for precise prostate segmentation via online sampling

Keyphrases

• Ablation Study 33%
• Branch Architecture 33%
• Conventional Learning 33%
• Cross-entropy Loss 66%
• CT Image Dataset 33%
• CT Images 33%
• Deep Metric Learning 33%
• Dice Loss 66%
• Dual-branch 33%
• Feature Map 33%
• Feature Space 33%
• Fully Convolutional Network 66%
• Ground-truth Labels 33%
• Image Level 33%
• Intermediate Features 33%
• Learned Features 33%
• Learning Methods 66%
• Learning Module 33%
• Metric Learning 66%
• Metric Loss 33%
• Multi-task 33%
• Multi-task Learning 33%
• Multi-task Network 33%
• Network Architecture 33%
• Nonsmooth 33%
• Online Metric Learning 33%
• Online Sampling 100%
• Prostate 66%
• Prostate Segmentation 100%
• Semantic Segmentation 33%
• State-of-the-art Techniques 33%
• Tissue Contrast 33%
• Training Phase 33%
• Triplet 66%
• Tuple 66%
• Two-stage Framework 33%
• U-Net 33%
• U-Net Architecture 33%
• V-Net 33%
• Voxel-wise 66%

Computer Science

• Conventional Learning 33%
• Feature Map 33%
• Feature Space 33%
• Fully Convolutional Network 66%
• Image Segmentation 33%
• Learned Feature 33%
• Multitask Learning 33%
• Network Architecture 33%
• Subnetwork 100%
• Training Phase 33%