def test_distance_horizontal():
"""
Test horizontal points, second point is right
to first point
"""
p1 = (0, 0)
p2 = (0, 1)
assert distance(p1, p2) == 1
def test_distance_vertical():
"""
Test vertical points, second point is above
first point
"""
p1 = (0, 0)
p2 = (1, 0)
assert distance(p1, p2) == 1
def test_distance_vertical_reverse():
"""
Test vertical points, second point is below
first point
"""
p1 = (1, 0)
p2 = (0, 0)
assert distance(p1, p2) == 1
def spatial_colocalize(fileinfos, options: Options):
"""Performs spatial colocalization of the given files
Arguments:
fileinfos {iterable of ChannelInfo} -- [description]
options {Options} -- [description]
"""
assert _currentrun, "please call new_run before starting a analysis"
data = list()
for info in fileinfos:
file, datafile, maskoutput, channelname = info
_log("Loading file \"{0}\"".format(file))
image = imread(file)
dataimg = imread(datafile) if datafile else None
data.append(
# todo: turn into named touple?
(image, dataimg, maskoutput, channelname))
images = next(zip(*data)) # first element of each tuple is the image
# these are not the final masks!
masks = [image > options.threshold for image in images]
connectivity = 2 if options.diagnoalconnectivity else 1
labeled_masks = [
skimage.measure.label(m, connectivity=connectivity) for m in masks
]
overlapping_areas = (np.ones_like(masks[0]) == 1) # Full true array
for mask in masks:
overlapping_areas = np.logical_and(overlapping_areas, mask)
selection_masks = list()
for index in range(0, len(masks)):
ids = np.unique(labeled_masks[index][overlapping_areas])
_log("{0} objects in \"{1}\"".format(len(ids), fileinfos[index][3]))
selection_mask = np.in1d(labeled_masks[index],
ids).reshape(images[index].shape)
# selections masks selects features, that overlap with other features
mask_filename = data[index][2]
if mask_filename: # third entry of the tuple contains a filename, if mask should be saved
_log("Saving mask as {0}".format(mask_filename))
save_binary_image(mask_filename, selection_mask)
selection_masks.append(selection_mask)
# selection mask labels
mask_labels = [
skimage.measure.label(mask, connectivity=connectivity)
for mask in selection_masks
]
if options.csvfolder:
_log("Calculating statistics")
os.makedirs(os.path.dirname(options.csvfolder), exist_ok=True)
standard_csv_lines = list()
overlap_csv_lines = list()
for index in range(0, len(masks)):
_log("Working on statistics from {0}".format(data[index][3]))
current_labels = mask_labels[index]
regionprops = skimage.measure.regionprops(
current_labels, intensity_image=data[index][1], cache=False)
for region in regionprops:
csvrow = dict()
csvrow["channel"] = fileinfos[index][3]
csvrow["object_id"] = region.label
csvrow["sourcefile"] = fileinfos[index][0]
if "area_px" in options.statistics:
#_log("Calculating particle areas")
csvrow["area_px"] = region.area
if "intensity_mean" in options.statistics:
#_log("Calculating intensity avg")
try:
csvrow["intensity_mean"] = region.mean_intensity
except AttributeError:
_log(
"Trying to calculate intensity without data file!")
csvrow["intensity_mean"] = "No data image specified!"
if "intensity_max" in options.statistics:
#_log("Calculating intensity max")
try:
csvrow["intensity_min"] = region.min_intensity
csvrow["intensity_max"] = region.max_intensity
except AttributeError:
_log(
"Trying to calculate intensity without data file!")
csvrow["intensity_min"] = "No data image specified!"
csvrow["intensity_max"] = "No data image specified!"
if "com" in options.statistics:
csvrow["com_unweighted"] = region.centroid
csvrow["com_unweighted_local"] = region.local_centroid
try:
csvrow["com_weighted"] = region.weighted_centroid
csvrow[
"com_weighted_local"] = region.weighted_local_centroid
except AttributeError:
pass # no data file specified
if "area_overlap_px" in options.statistics:
# Area overlapping in all channels
minimal_overlapping_area = 0
# todo: filter reverse overlaps
for comparisonchannel in [
i for i in range(0, len(masks)) if i != index
]:
own_id = region.label
own_selection = mask_labels[index] == own_id
comparison_ids = np.array(
labeled_masks[comparisonchannel])
print(comparison_ids)
comparison_ids[np.logical_not(own_selection)] = 0
# comparison_ids now contains ids present in both channels
ids_to_select = np.unique(comparison_ids)
# now create the final label mask, since the above solution cuts of
# areas not in both channels!
comparison_label_mask = np.in1d(
labeled_masks[comparisonchannel],
ids_to_select).reshape(images[index].shape)
comparison_labels = np.array(
labeled_masks[comparisonchannel])
comparison_labels[np.logical_not(
comparison_label_mask)] = 0
if DEBUG:
print(comparison_labels)
# todo: implement data image
overlappingregions = None
if data[comparisonchannel][1]:
overlappingregions = skimage.measure.regionprops(
comparison_labels,
intensity_image=data[comparisonchannel][1],
cache=False)
else:
overlappingregions = skimage.measure.regionprops(
comparison_labels, cache=False)
cumulative_overlap = 0
if DEBUG:
print("Comparing")
print(comparison_label_mask)
minimal_overlapping_area = np.logical_and(
comparison_label_mask, overlapping_areas)
minimal_overlapping_size = np.count_nonzero(
minimal_overlapping_area)
for overlap in overlappingregions:
detailedoverlapcsvrow = dict()
detailedoverlapcsvrow["source_channel"] = data[
index][3]
detailedoverlapcsvrow["source_id"] = region.label
detailedoverlapcsvrow["comparison_channel"] = data[
comparisonchannel][3]
detailedoverlapcsvrow[
"comparison_id"] = overlap.label
detailedoverlapcsvrow[
"area_in_both_channels"] = overlap.area
# calculate distances
distance_unweighted = distance(
region.centroid, overlap.centroid)
angle_unweighted = angle(region.centroid,
overlap.centroid)
distance_weighted = "No data file specified"
angle_weighted = "No data file specified"
detailedoverlapcsvrow[
"distance_unweighted"] = distance_unweighted
detailedoverlapcsvrow[
"angle_unweighted"] = angle_unweighted
try:
distance_weighted = distance(
region.weighted_centroid,
overlap.weighted_centroid)
angle_weighted = angle(
region.weighted_centroid,
overlap.weighted_centroid)
except AttributeError:
pass
detailedoverlapcsvrow[
"distance_weighted"] = distance_weighted
detailedoverlapcsvrow[
"angle_weighted"] = angle_weighted
cumulative_overlap += overlap.area
overlap_csv_lines.append(detailedoverlapcsvrow)
csvrow["cumulative_overlap_{0}_{1}".format(
data[index][3],
data[comparisonchannel][3])] = cumulative_overlap
csvrow[
"minimum_overlapping_area"] = minimal_overlapping_size
standard_csv_lines.append(csvrow)
# finally: save csv data
standardfile = os.path.join(
options.csvfolder,
"statistics_{0:%Y-%m-%d %H-%M-%S}.csv".format(_currentrun))
overlapfile = os.path.join(
options.csvfolder,
"overlaps_{0:%Y-%m-%d %H-%M-%S}.csv".format(_currentrun))
_writestats(standardfile, standard_csv_lines)
_writestats(overlapfile, overlap_csv_lines)