Segmenting mother machine data with Omnipose

This notebook is adapted from the Omnipose docs’ basic tutorial which can be found here: https://omnipose.readthedocs.io/examples/mono_channel_bact.html

Here we simply load a single TIFF stack of single mother machine trenches, and tile the images for more efficient segmentation.

import numpy as np
from cellpose import models, core
from cellpose import plot
import omnipose
from cellpose import models
from glob import glob
from natsort import natsorted

# This checks to see if you have set up your GPU properly.
# CPU performance is a lot slower, but not a problem if you
# are only processing a few images.
use_GPU = core.use_gpu()
print('>>> GPU activated? %d'%use_GPU)

import tifffile
from skimage.transform import rescale, resize, downscale_local_mean

# for plotting
import matplotlib as mpl
import matplotlib.pyplot as plt

2022-08-25 20:43:08,194 [INFO] ** TORCH CUDA version installed and working. **
>>> GPU activated? 1

Load in the first 8 frames of the timeseries data

real_data = tifffile.imread("real_data/trench_0.tif")[:8]
real_data = np.concatenate([data for data in real_data], axis=1)

Use omnipose’s preprocessing step.

from cellpose import io, transforms
from omnipose.utils import normalize99
imgs = [real_data]
nimg = len(imgs)
fig = plt.figure(figsize=[40]*2) # initialize figure
for k in range(len(imgs)):
    img = transforms.move_min_dim(imgs[k]) # move the channel dimension last
    if len(img.shape)>2:
        imgs[k] = np.mean(img,axis=-1) # or just turn into grayscale

    imgs[k] = normalize99(imgs[k])
    plt.subplot(1,len(imgs),k+1)
    plt.imshow(imgs[k],cmap='gray')
    plt.axis('off')

Load in the last model in the model directory in the training data directory.

model_list = natsorted(glob("omnipose_training_data_large/models/*"))
model_name = model_list[-1]
print(model_name)
use_gpu = use_GPU# = False
model = models.CellposeModel(gpu=use_gpu, pretrained_model=model_name, omni=True, concatenation=True)

omnipose_training_data_large/models/cellpose_residual_on_style_on_concatenation_off_omni_omnipose_training_data_large_2022_08_17_19_22_08.661299_epoch_3999
2022-08-25 20:43:08,897 [INFO] ** TORCH CUDA version installed and working. **
2022-08-25 20:43:08,897 [INFO] >>>> using GPU

Segment the image

chans = [0,0] #this means segment based on first channel, no second channel

n = [0] # make a list of integers to select which images you want to segment
n = range(nimg) # or just segment them all

# define parameters
mask_threshold = -1
verbose = 0 # turn on if you want to see more output
use_gpu = use_GPU #defined above
transparency = True # transparency in flow output
rescale=None # give this a number if you need to upscale or downscale your images
omni = True # we can turn off Omnipose mask reconstruction, not advised
flow_threshold = 0. # default is .4, but only needed if there are spurious masks to clean up; slows down output
resample = True #whether or not to run dynamics on rescaled grid or original grid
masks, flows, styles = model.eval([imgs[i] for i in n],channels=chans,rescale=rescale,mask_threshold=mask_threshold,transparency=transparency,
                                  flow_threshold=flow_threshold,omni=omni,resample=resample,verbose=verbose)

Visualise the segmentation with omnipose’s visualisation.

for idx,i in enumerate(n):

    maski = masks[idx]
    flowi = flows[idx][0]

    fig = plt.figure(figsize=(10,12))
    plot.show_segmentation(fig, omnipose.utils.normalize99(imgs[i]), maski, flowi, channels=chans, omni=True, bg_color=0)

    plt.tight_layout()
    plt.savefig(f"omnipose_diagnostics/{str(idx).zfill(4)}.png", dpi = 120)
    plt.show()