def apply_extract_features_cp(well_tile,filepattern):
wildcards = {'well':well_tile[0],'tile':well_tile[1]}
filepattern.update(wildcards)
stacked = read_hdf_image(name(filepattern))
nuclei = read(name(filepattern,subdir='process_ph',tag='nuclei',ext='tif'))
cells = read(name(filepattern,subdir='process_ph',tag='cells',ext='tif'))
df_result = Snake._extract_phenotype_cp(data_phenotype=stacked,
nuclei=nuclei,
cells=cells,
wildcards=wildcards,
nucleus_channels=[0,1,2,3],
cell_channels=[0,1,2,3],
channel_names=['dapi','tubulin','gh2ax','phalloidin']
)
df_result.to_csv(name(filepattern,subdir='process_ph',tag='cp_phenotype',ext='csv'))
def calculate_illumination_correction(files, smooth=None, rescale=True, threading=False, slicer=slice(None)):
"""calculate illumination correction field for use with apply_illumination_correction
Snake method. Equivalent to CellProfiler's CorrectIlluminationCalculate module with
option "Regular", "All", "Median Filter"
Note: algorithm originally benchmarked using ~250 images per plate to calculate plate-wise
illumination correction functions (Singh et al. J Microscopy, 256(3):231-236, 2014)
"""
from ops.io import read_stack as read
N = len(files)
global data
data = read(files[0])[slicer]/N
def accumulate_image(file):
global data
data += read(file)[slicer]/N
if threading:
from joblib import Parallel, delayed
Parallel(n_jobs=-1,require='sharedmem')(delayed(accumulate_image)(file) for file in files[1:])
else:
for file in files[1:]:
accumulate_image(file)
data = np.squeeze(data.astype(np.uint16))
if not smooth:
# default is 1/20th area of image
# smooth = (np.array(data.shape[-2:])/8).mean().astype(int)
smooth = int(np.sqrt((data.shape[-1]*data.shape[-2])/(np.pi*20)))
selem = skimage.morphology.disk(smooth)
median_filter = ops.utils.applyIJ(skimage.filters.median)
with warnings.catch_warnings():
warnings.simplefilter("ignore")
smoothed = median_filter(data,selem,behavior='rank')
if rescale:
@ops.utils.applyIJ
def rescale_channels(data):
# use 2nd percentile for robust minimum
robust_min = np.quantile(data.reshape(-1),q=0.02)
robust_min = 1 if robust_min == 0 else robust_min
data = data/robust_min
data[data<1] = 1
return data
smoothed = rescale_channels(smoothed)
return smoothed
def accumulate_image(file):
global data
data += read(file)[slicer]/N
.askopenfilename(initialdir=dirname(label_file),
title="Select intensity image (.tif)",
filetypes=(("image files","*.tif"),
("all files","*.*")
)
)
)
root_2.destroy()
_,ext_2= splitext(filename_2)
if ext_2 not in image_ext:
raise Exception('intensity image file must be .tif or .tiff')
else:
img_file = filename_2
img_labels = read(label_file)
df = (feature_table(img_labels, img_labels, features_basic)
.assign(label_file=label_file,img_file=img_file)
)
else:
table_file = filename_1
raise Exception('input table file must be .csv or .hdf')
elif len(sys.argv) == 2:
filename_1 = sys.argv[1]
_,ext_1= splitext(filename_1)
if ext_1 not in table_ext:
raise Exception('input table file must be .csv or .hdf')
else: