hisaylama / misic

segmentation of rod-shaped bacteria

MiSiC

Microbe segmentation in dense colonies.

Installation

Requires version python version 3.6/7

pip install misic

MiSiC as napari plugin

A napari plugin is up and running at the moment with a few glitches!

Here are the steps for installation:

1. You can download napari at the bottom of the page at https://github.com/napari/napari/releases
2. After you install napari, in the plugins tab > Install plugin > misic-napari
3. Restart napari
4. Drag and drop an image
5. Go to segment More information is at https://pypi.org/project/misic-napari/

The plugin is still underdevelopment so there might be a few bugs that will be incrementally updated.

Possible errors during installtion

tensorflow dependecies may not be installed if pyparsing error pip install pyparsing if protobuf error conda install protobuf

Usage

use package

from misic.misic import *
from misic.extras import *
from skimage.io import imsave,imread
from skimage.transform import resize,rescale

filename = 'awesome_image.tif'

# read image using your favorite package
im = imread(filename)
sr,sc = im.shape

# Parameters that need to be changed
## Ideally, use a single image to fine tune two parameters : mean_width and noise_variance (optional)

#input the approximate mean width of microbe under consideration
standard_width = 9.7

# the approximate width of cells to be segmented
mean_width = 9.7

# If image is phase contrast light_background = True
light_background = True

# compute scaling factor
scale = (standard_width/mean_width)

# Initialize MiSiC
mseg = MiSiC()

# preprocess using inbuit function or if you are feeling lucky use your own preprocessing
# recomended preprcessing
# im = adjust_gamma(im,0.25)
# im = unsharp_mask(im,2.2,0.6)

# for fluorescence images
# im = gaussian(laplace(im),2.2)
# im = add_noise(im,0.1)
# OR
# im = random_noise(im,mode = 'gaussian',var = 0.1/100.0)

im = rescale(im,scale,preserve_range = True)

# add local noise
img = add_noise(im,sensitivity = 0.13,invert = light_background)

# segment
yp = mseg.segment(img,invert = light_background)
yp = resize(yp,(sr,sc))

# watershed based post processing (optional)
# yp = postprocess_ws(img,yp)
yp = postprocessing(im if light_background else -im,yp)[:,:,0]  

# save 8-bit segmented image and use it as you like
imsave('segmented.tif', ((yp > 0)*255).astype(np.uint8))
''''

### In case of gpu error, one might need to disabple gpu before importing MiSiC [ os.environ["CUDA_VISIBLE_DEVICES"]="-1" ]