Seedings results comparison

Key Investigators

• Harvard Medical School: Antonin Perrot-Audet, Kishore Mosaliganti, Sean Megason
• RPI: Badri Roysam, Raghav Padmanabhan

Project

Objective

• Improve segmentation algorithms initialization for nuclei detection in 3D fluorescent microscopy:
  • get a better accuracy,
  • improve computation speed.

Approach, Plan

• Compare different algorithms,
• Find a measure to evaluate different algorithm results,
• Fusion output of several algorithms.

Progress

• Implemented 3 different algorithms in various langages:
  • Radial voting,
  • Hough transform,
  • Multi-scale Distance Map weighted Laplacian of Gaussian.
• Created a collaboration framework using ITK:
  • Set of utilities for input images normalization
  • Developed a windowed local maxima filter
• Evaluated and compared output of on synthetic 2D data & 3D Nuclei channel from Megason Lab
  • Multi-scale Distance Map weighted Laplacian of Gaussian.
  • Radial voting

Ressources

• Source code :

git@github.com:antonin07130/NAMICSeeding.git

We shall use git for version control : A small introduction to git : here

• Data set :

Non public