Python3 scripts for analyzing calcium imaging signals.
Registration (landmark based translation), cell detection, and parallel clustering (k-means) on the movies are implemented.
numpy, scipy, scikit-learn, scikit-image, nibabel
ipython [optional]
On Ubuntu (tested under 16.04)
sudo apt-get install python3 python3-numpy python3-scipy
sudo apt-get install python3-sklearn python3-skimage python3-nibabel
sudo apt-get install python3-nibabel
On Windows (not tested)
Download installers from the sites, and follow the instructions.
class preprocessor.Movie
| Name | Description |
|---|---|
| New | loads movie from a file |
| to_nifti1 | saves movie as nifti 1 image |
| to_nifti2 | saves movie as nifti 2 image |
| get_data | gets the internal movie array (tzyx order) |
class preprocessor.MovieMapper
| Name | Description |
|---|---|
| set_landmarks_properties | sets shape of the landmarks |
| set_target_time_range | sets time range |
| get_map | creates a MovieMap class of the movie |
class preprocessor.MovieMap
| Name | Description |
|---|---|
| get_mapped | creates a mapped movie (Movie class) |
class cluster.KMeans
The interface is almost identical to scikit-learn KMeans class.
Registration on a movie.
$ ipython3
In [0]: from canal.preprocessor import Movie, MovieMap, MovieMapper # import classes
In [1]: movie = Movie.New('sample.nii') # read nifti movie
In [2]: mapper = MovieMapper()
In [3]: mapper.set_landmark_properties((128, 128)) # landmark shape set to (128, 128)
In [4]: mapper.set_target_time_range((0, 50)) # take median of the timeframes in 0:50
In [5]: map = mapper.get_map(movie) # get MovieMap of the movie (take your time)
In [6]: reg_movie = map.get_mapped(movie) # create movie applied the registration (take a coffee break)
Saving the movie (nifti 1 or 2 format).
In [7]: affine = np.eye(4)
In [8]: reg_movie.to_nifti1('reg_sample.nii', affine)
Cell detection.
In [9]: from canal.cell import find_cell
In [10]: data = reg_movie.get_data() # tzyx order numpy array
In [11]: volume = data.mean(axis = 0) # make a 3d volume
In [12]: cell_centers = find_cell(volume, 4) # find cells assuming their radii are 4