search

AllenInstitute / neuron_morphology

Tools for working with single-neuron morphological reconstructions
https://neuron-morphology.readthedocs.io/en/latest/
Other
19 stars 8 forks source link

EM transform and features #184

Open wbwakeman opened 4 years ago

wbwakeman commented 4 years ago
  1. Work with Forrest to identify feature set that is appropriate for EM data
  2. Implement this as a "feature group" in neuron_morphology
  3. Work with them to understand the transform requirements and how they can be implemented to make a reconstruction suitable for computing features on

Forrest and Nuno write:

Regarding the transform, my understanding is that neuron morphology requires a separate file that indicates the locations of the layer borders in the y coordinate.

The pia “up” direction in pinky is 0.12253594, 0.95501504, 0.27005781. So essentially aligned along the y axis. Then the layer borders need to be estimated, we know that the top of the volume is at the layer 1/ layer 2/3 border, and can estimate the other layer borders from there. .but the truth is most of the neuron_morphology features are not going to make a lot of sense on this dataset because the axons/dendrites are so cutoff. I’m happy to work on providing them though because working out these kinds of logistics is precisely the valuable part of going through this exercise.

Regarding the layer borders, apart from the top of the reconstructions being at the layers 1/2 borders as Forrest described bellow there is no other border that can be added. The bottom of the reconstructions to not reach layer 4, which would be the next reasonable layer border to identify