Trajectory curvature

Hi all,

I have mice trajectories that I was able to categorize visually in 3 groups as followed :

I’m now trying to find a way to make this classification automatic by finding a measurement that would allow splitting into these categories. However I can’t find the right metric to do so.

As I’m not familiar with behaviour studies, would anyone have a good idea about what metrics to use ? Unfortunately, I have to add this measurement in Matlab but already having some pointers into what I should search would help.

Thanks a lot !

Hello Laurent,

I would try to measure a bunch of numerical features on individual tracks like

  • max distance traveled
  • confinement ratio
  • mean straight line speed
  • linearity of forward progression
    etc.
    and use them in simple decision tree classifier.

I tried that a while ago to investigate Helicobacter pylori mutants and it worked well enough.

2 Likes

Hi @tinevez

Thanks for your reply ! I’ll try and play with those settings and come back and marked as solved if this worked.

I might also come with new questions :wink: