Use additional information to resolve tracks?

[nodice73]

Hello,

I find trackpy to be intuitive and a pleasure to use. Thank you for making a great module!

I was wondering if there was a way to include more information than position to help the algorithm resolve close particles. For example, I have cell data which includes e.g., their area, fluorescence intensity, etc. I've noticed mistakes in high-density areas that would be easily resolved if there was some concept of weighting potential links by the value(s) of some other parameter(s).

I'm guessing this would be similar to how velocity prediction is implemented.

Thanks again!

[nkeim]

Hi! Thanks for using trackpy! I’m not quite sure what you mean, but since trackpy supports n-dimensional tracking, you can use quantities like intensity or size to assist linking. You just add it to the list you specify with pos_columns. You usually need to rescale the quantity so that differences in size values (for example) are comparable to distances in space; you may also need to increase search_range. I have done this before with some success. On Jan 10, 2018, 6:06 PM -0800, Adam Waite <notifications@github.com>, wrote: Hello, I find trackpy to be intuitive and a pleasure to use. Thank you for making a great module! I was wondering if there was a way to include more information than position to help the algorithm resolve close particles. For example, I have cell data which includes e.g., their area, fluorescence intensity, etc. I've noticed mistakes in high-density areas that would be easily resolved if there was some concept of weighting potential links by the value(s) of some other parameter(s). I'm guessing this would be similar to how velocity prediction is implemented. Thanks again! — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub<#470>, or mute the thread<https://github.com/notifications/unsubscribe-auth/AC2NbSERlYCJ6mLj-AU6xr2EABaxad9Hks5tJWyRgaJpZM4RaNzz>.

[caspervdw]

To add to @nkeim: you can also specify the search_range per feature (x, y, intensity, ...) simply by passing search_range=(7, 7, 100) to link. This will rescale the features for you and basically works the same as Nathan’s suggestion.

I am not sure if this feature is in the latest release, but it is for sure present in the current master version of the code.

[nodice73]

Yes, this is exactly what I was talking about.

You might want to mention this explicitly in the docs, as I was being too literal about the meaning of 'pos_columns'.

Thanks again!