/HE-tumor-object-segmentation

Segmentation of tumor objects in H&E images using graph-based clustering based on opponent color space and internal image statistics

Primary LanguageC++

HE-tumor-object-segmentation

Segmentation of tumor objects in H&E images using graph-based clustering based on opponent color space and internal image statistics Author: Luong Nguyen Email: lun5@pitt.edu Started 10/23/2014

Main components:

  1. Inputs: tiling of WSI, parallelization, rotation matrix
  2. Annotations: collect annotations from pathologists
  3. Affinity calculation: Features and affinity function
  4. Clustering: NMinCut, Embedded Spectral Clustering

Reused codes and packages from Citation of KDE Citation of Multiscale Spectral Clustering Citation of Crisp Boundary