NeuroXplorer

Introductory Gala Watershed Process Documentation

Augusto Ramirez, Thomas Keady, Albert Lee

Part 1 : Data collection

Data collection tutorials can be found at https://github.com/openconnectome/ndio-demos.git. Detailed documentation also exists at https://github.com/openconnectome/ndio. Data collected from ndio and saved as a numpy file with the name 'nXp_data.npy' can be run through our watershed process when placed in the same folder .

The data used for our watershed process can be found at https://www.dropbox.com/s/95rsvfdhkiodq1x/nXp_data.npy?dl=0. The rest of the necessary programs are located on our github repo at https://github.com/Connectomics-Classes/NeuroXplorer to compare results.

Part 2 : Watershed

The complete watershed file can be found in our github repo in the folder 'testing' under the name 'gala_shed.py'. As long as the 'nxp_data.np' file is in the same folder, the program can be immediately run from terminal using the command 'python gala_shed.py'. The following steps deconstruct the program on a very basic level.

Step 1:

First launch python and import the following packages. If an error states that a certain module does not exist simply use 'pip install insert package name here' in order to download the package. Detailed instructions on using pip can be found at https://pip.pypa.io/en/stable/installing/.



In [ ]:

    
### Watershed segmentation imports
from gala import classify, morpho, features, agglo, evaluate as ev, optimized 
import numpy as np
from skimage.morphology import closing
from skimage.morphology import binary_dilation
from skimage.morphology import binary_erosion
from skimage.filters.rank import threshold
import matplotlib.pyplot as plt
###

print "Done importing packages"

Step 2

Next load the necessary data from our 'nxp_data.npy file'. If you are using your own data, make sure you have already replaced our npy file with your own. In any case, please make sure the file you are using is named 'nxp_data.np'.



In [ ]:

    
# Gather images from nXp_data.npy file
print "Getting images"
membrane_images = np.load('nXp_data.npy')

# Assigning one image from array to variable pre_image
pre_image = 255 * membrane_images[0]

min_seed_size = 5 #2
connectivity = 2 #2
smooth_thresh = 0.02 #0.02

Step 3

The below code is for the actual watershed segmentation. The distance parameters can be modified to produce different results.



In [ ]:

    
########
# Watershed segmentation
########
print "Running watershed"

ws_train = morpho.watershed(pre_image,
connectivity=connectivity, smooth_thresh=smooth_thresh,
override_skimage=1,minimum_seed_size=min_seed_size) + 1
#ws_train = ws_train

fig, axes = plt.subplots(ncols=1, figsize=(8, 2.7), sharex=True, sharey=True, subplot_kw={'adjustable':'box-forced'})
ax0 = axes

ax0.imshow(ws_train)#, cmap=plt.cm.spectral, interpolation='nearest')
ax0.set_title('Overlapping objects')

Step 5

Finally a plot of the annotated image should pop up.



In [ ]:

    
fig.subplots_adjust(hspace=0.01, wspace=0.01, top=0.9, bottom=0, left=0,
                    right=1)

plt.show()