Title: An Unsupervised Learning Model for Deformable Medical Image Registration Paper Link
Authors: Guha Balakrishnan et al.
Association: MIT
Submission: 2018
…………………………………………………………………………………………………………………
»> Background of Deep Learning in Medical Image Analysis
»> A Review: Robot-Assisted Endovascular Catheterization Technologies:
…………………………………………………………………………………………………………………
»> Image Registration Basic Knowledge
»> Image Registration Literature Review
»> Slice-To-Volume Medical Image Registration Background
…………………………………………………………………………………………………………………
Motivation
Contributions
(0) define registration as a parametric function, and optimize its parameters given a set of images from a collection of interest. Given
(1) (OPEN SOURCE) code is available at https://github.com/balakg/voxelmorph
(2) present a learning-based solution requiring no supervised information such as ground truth correspon- dences or anatomical landmarks during training,
(3) propose a CNN function with parameters shared across a population, enabling registration to be achieved through a function evaluation
(4) enables parameter optimization for a variety of cost functions, which can be adapted to various tasks.
Methods
Loss function
consisting of two components:
- penalizes differences in appearance L_sim
And set L_sim to the negative local cross-correlation of M(φ) and F.
- penalizes local spatial variations in φ L_smooth
Performance
