Research

GPU-BASED RECONSTRUCTION

3D reconstruction

  • Reconstruction of 3D volumetric image from projection data acquired CT
  • Very time-consuming task
  • Accelerated by more than 50 times using GPU
METAL ARTIFACT REDUCTION

Metal Artifact Reduction

  • Streak artifacts as dark and bright streaks in CT
  • Exact metal segmentation and proper inpainting
 Volume Visualization

Transfer Functio­n

Industrial Volume Visualization

  • Very large data (1GB~)
  •  Metal artifact reduction

­                    

  • Defect detection

Technical feature

  • Various visualization and acceleration techniques
  • GPU shader model
 ­ 

Mesh etraction & Simplification

Visualization Miture

Reverse engineering and analysis on industrial CT volume

Enhanced 3D Volume Rendering

Occlusion Map

  • 2D Transfer Function: intensity vs. ambient occulusion

Contet-Preserving VR

  • Simultaneously visualize surface and internal structures
  • Modulate the opacity based on features of each sample

MIDA

  • Alter the accumulation behavior of DVR to exhibit characteristics similar to MIP
  • Preserve local maxima

­                    

 High Quality Volume Rendering

Virtual sampling

  • Assume that the transition within a ray segment is like curve
  • Virtual sampling occurs with the given ray segment via curve interpolation

Medical Image Processin­­g

Liver vessel segmentation and hepatic/portal vein classification

  • Liver vessel segmentation and hepatic/portal vein classification
  • Nonrigid registration between multi-phase CTAs

 

  

Object segmentation via deep convolutional neural network (CNN)

  • Object shape prediction by 3D deconvolutional network
  • Image block-wise prediction with CNN classifier
Non-rigid registration for lung CT