We propose a novel attention gate (AG) model for medical imaging that automatically learns to focus on target structures of varying shapes and sizes.
SOTA for Pancreas Segmentation on CT-150
Implementation of different kinds of Unet Models for Image Segmentation - Unet , RCNN-Unet, Attention Unet, RCNN-Attention Unet, Nested Unet
There is large consent that successful training of deep networks requires many thousand annotated training samples.
Proposed CNN based segmentation approaches demonstrate how 2D segmentation using prior slices can provide similar results to 3D segmentation while maintaining good continuity in the 3D dimension and improved speed.