CUDA Spatial Deform by Qingsong Yao

A fast tool to do image augmentation by CUDA on GPU(especially elastic deformation), can be helpful to research on Medical Image Analysis.      https://github.com/qsyao/cuda_spatial_deform

Authors：Qingsong Yao worked as an intern in system group of Microsoft research Asia for one year, practicing CUDA programming.

Motivation

  1. When the size of image is too large, it takes a lot of time(much more than forward and backward computation say in U_Net), especially for 3D image(like CT).
  2. Elastic deformation on CPU is too slow.
  3. Doing Sptial_Deform by muti-processing consumes of too much CPU resources, to which most GPU servers(like 32 cores with 4 gpus) can not afford.

Implementation Overview

  1. Doing Spation_Deform on GPU instead of CPU, greatly saving CPU resources.
  2. Very Fast, speed up 25x for rotation, 45x for elastic_deformation.
  3. Support many types of spatial deform: flip, rotate, scale, translate, elastic_deformation.
  4. Support many rules of map_coordinates: mirror, constant, reflect, wrap, nearest.
  5. Doing Spatial_Deform by doing calculations of coordinates, all transformations get combined before they are applied to the image
  6. Implement map_coordinates by linear interpolation(for image) and the nearest interpolation(for labels).
  7. Users can fetch coordinates from CUDA and do cubic interpolation at CPU by scipy.map_coordinates(order = 3)

Citation

If you use our code, please cite our paper:

Chao Huang, Hu Han, Qingsong Yao, Shankuan Zhu, S. Kevin Zhou. , 3D U2-Net: A 3D Universal U-Net for Multi-Domain Medical Image Segmentation, MICCAI 2019.