def train_texture_samples(subdir, samples_dict, regression_variable, endings, samples_per_image, age_ranges):
'''
Read in samples for texture classification.
'''
# Organize some information.
total_files = os.listdir(subdir)
data_points = [' '.join(a_file.split('.')[0].split(' ')[0:-1]) for a_file in total_files if a_file.split(' ')[-1] == 'hmask.png']
with open(subdir + os.path.sep + 'position_metadata_extended.json', 'r') as read_file:
my_metadata = json.loads(read_file.read())
metadata_dict = {my_metadata[i]['timepoint']: my_metadata[i][regression_variable.lower()] for i in range(0, len(my_metadata))}
egg_age_dict = {my_metadata[i]['timepoint']: my_metadata[i]['egg_age'] for i in range(0, len(my_metadata))}
# Exclude larval animals from ghost_age computation.
if regression_variable.lower() == 'ghost_age':
data_points = [a_point for a_point in data_points if egg_age_dict[a_point] > 0]
# Actualyl sample my images.
for a_point in data_points:
my_age = metadata_dict[a_point]/24
if regression_variable.lower() == 'ghost_age':
my_age = abs(my_age)
mask_file = freeimage.read(subdir + os.path.sep + a_point + ' ' + endings[1])
image_file = freeimage.read(subdir + os.path.sep + a_point + ' ' + endings[0])
my_samples = sample_8star(image_file, mask_file, 17, sample_number = samples_per_image)
for age_range in age_ranges:
if age_range[0] <= my_age < age_range[1]:
samples_dict[age_range].extend(my_samples)
return samples_dict
def measure(self, position_root, timepoint, annotations, before, after):
derived_root = position_root.parent / DERIVED_ROOT
image_file = position_root / f'{timepoint} {self.image_type}.png'
if not image_file.exists():
return [numpy.nan] * len(self.feature_names)
image = freeimage.read(image_file)
flatfield = freeimage.read(position_root.parent / 'calibrations' /
f'{timepoint} fl_flatfield.tiff')
image = image.astype(numpy.float32) * flatfield
mask = self.get_mask(position_root, derived_root, timepoint,
annotations, image.shape)
if mask is None:
return [numpy.nan] * len(self.feature_names)
if mask.sum() == 0:
print(
f'No worm region defined for {position_root.name} at {timepoint}'
)
data, region_masks = measure_fluor.subregion_measures(image, mask)
if self.write_masks:
color_mask = measure_fluor.colorize_masks(mask, region_masks)
out_dir = derived_root / 'fluor_region_masks' / position_root.name
out_dir.mkdir(parents=True, exist_ok=True)
freeimage.write(color_mask,
out_dir / f'{timepoint} {self.image_type}.png')
return data
def read_corrected_gfp_fluorescence(gfp_img,
calibration_directory,
super_vignette,
hot_threshold=10000):
'''
Correct fluorescence images for flatfield, and re-normalize to make the images more nicely viewable.
'''
# Read in image and apply vignetting.
gfp_img = pathlib.Path(gfp_img)
timepoint = gfp_img.name.split(' ')[0]
super_vignette = pickle.load(open(super_vignette, 'rb'))
raw_image = freeimage.read(gfp_img)
raw_image[np.invert(super_vignette)] = 0
calibration_directory = pathlib.Path(calibration_directory)
# Correct for flatfield.
flatfield_path = calibration_directory.joinpath(timepoint +
' fl_flatfield.tiff')
calibration_image = freeimage.read(flatfield_path)
#corrected_image = raw_image
corrected_image = raw_image * calibration_image
# Correct for hot pixels.
median_image = ndimage.filters.median_filter(corrected_image, size=3)
difference_image = np.abs(
corrected_image.astype('float64') -
median_image.astype('float64')).astype('uint16')
hot_pixels = difference_image > hot_threshold
median_image_hot_pixels = median_image[hot_pixels]
corrected_image[hot_pixels] = median_image_hot_pixels
# Return the actual image.
return corrected_image.astype(np.uint16)
def extract_image_set(image_files, out_dir, date, age, plate_params, ignore_previous=False):
"""Find wells in a set of scanner images and extract each well into a separate image
for further processing.
Parameters:
image_files: list of paths to a set of images.
out_dir: path to write out the extracted images and metadata.
date: date object referring to image scan date
age: age in days of the worms in these images
plate_params: configuration information for extracting wells from the plates.
This must be a parameter dictionary suitable to pass to extract_wells.extract_wells()
ignore_previous: if False, and stored results already exist, skip processing
"""
out_dir = pathlib.Path(out_dir)
metadata = out_dir / 'metadata.pickle'
if metadata.exists() and not ignore_previous:
return
images = []
print('extracting images for {}'.format(out_dir))
well_mask = freeimage.read(str(out_dir.parent / 'well_mask.png')) > 0
for image_file in image_files:
image = freeimage.read(image_file)
if image.dtype == numpy.uint16:
image = (image >> 8).astype(numpy.uint8)
images.append(image)
well_names, well_images, well_centroids = extract_wells.extract_wells(images, well_mask, **plate_params)
well_dir = util.get_dir(out_dir / 'well_images')
for well_name, well_image_set in zip(well_names, well_images):
for i, image in enumerate(well_image_set):
freeimage.write(image, str(well_dir/well_name)+'-{}.png'.format(i))
util.dump(metadata, date=date, age=age, well_names=well_names, well_centroids=well_centroids)
def archive_human_masks(human_directory, new_directory, work_directory):
'''
For a directory of hand-drawn masks, mask out everything in the accompanying bright-field file except for the worm itself and a 100-pixel surrounding area to save disk space. Also, re-compress all images to maximize compression and space efficiency.
'''
for a_subdir in os.listdir(human_directory):
if os.path.isdir(human_directory + os.path.sep + a_subdir):
folderStuff.ensure_folder(new_directory + os.path.sep + a_subdir)
for a_file in os.listdir(human_directory + os.path.sep + a_subdir):
if a_file.split(' ')[-1] == 'hmask.png':
if not os.path.isfile(new_directory + os.path.sep + a_subdir + os.path.sep + a_file):
print('Up to ' + a_subdir + ' ' + a_file + '.')
my_stem = a_file.split(' ')[0]
my_mask = freeimage.read(human_directory + os.path.sep + a_subdir + os.path.sep + my_stem + ' ' + 'hmask.png')
bf_path = human_directory + os.path.sep + a_subdir + os.path.sep + my_stem + ' ' + 'bf.png'
if os.path.isfile(bf_path):
my_image = freeimage.read(bf_path)
else:
my_image = freeimage.read(bf_path.replace(human_directory, work_directory))
area_mask = my_mask.copy().astype('bool')
distance_from_mask = scipy.ndimage.morphology.distance_transform_edt(np.invert(area_mask)).astype('uint16')
area_mask[distance_from_mask > 0] = True
area_mask[distance_from_mask > 100] = False
my_image[np.invert(area_mask)] = False
freeimage.write(my_image, new_directory + os.path.sep + a_subdir + os.path.sep + my_stem + ' ' + 'bf.png', flags = freeimage.IO_FLAGS.PNG_Z_BEST_COMPRESSION)
freeimage.write(my_mask, new_directory + os.path.sep + a_subdir + os.path.sep + my_stem + ' ' + 'hmask.png', flags = freeimage.IO_FLAGS.PNG_Z_BEST_COMPRESSION)
elif a_file.split('.')[-1] == 'json':
shutil.copyfile(human_directory + os.path.sep + a_subdir + os.path.sep + a_file, new_directory + os.path.sep + a_subdir + os.path.sep + a_file)
return
def score_image_set(out_dir, score_params, ignore_previous=False):
"""Score wells for a single day's scanned images.
Parameters:
out_dir: directory in which well_images directory is found, and to which score
data will be written.
score_params: configuration information for scoring wells for movement.
This must be a parameter dictionary suitable to pass to score_wells.score_wells()
ignore_previous: if False, and stored results already exist, skip processing
"""
out_dir = pathlib.Path(out_dir)
score_file = out_dir / 'scores.pickle'
if score_file.exists() and not ignore_previous:
return
print('scoring images for {}'.format(out_dir))
well_names = util.load(out_dir / 'metadata.pickle').well_names
well_mask = freeimage.read(str(out_dir.parent / 'well_mask.png')) > 0
well_dir = out_dir / 'well_images'
well_images = []
for well_name in well_names:
images = [freeimage.read(str(image)) for image in sorted(well_dir.glob(well_name+'-*.png'))]
well_images.append(images)
well_scores = score_wells.score_wells(well_images, well_mask, **score_params)
util.dump(score_file, well_names=well_names, well_scores=well_scores)
scores_out = [[name, str(score)] for name, score in zip(well_names, well_scores)]
util.dump_csv(scores_out, out_dir / 'scores.csv')
def get_calibrated_image(self,
well,
timepoint,
label,
calibration_type,
apply_vignette=True):
'''
well - string for well
timepoint - string for timepoint per YYYY-MM-DDtHHMM format
label - string suffix coming after timepoint
calibration_type - ['bf,'fl']
'''
image_fn = self.expt_path / well / (timepoint + ' ' + label + '.png')
calibration_fn = self.expt_path / 'calibrations' / (
timepoint + ' ' + calibration_type + '_flatfield.tiff')
img = freeimage.read(str(image_fn))
calibration_img = freeimage.read(str(calibration_fn))
ref_intensity = self.expt_mdata['brightfield metering'][timepoint][
'ref_intensity']
if (self.expt_path /
'super_vignette.pickle').exists() and apply_vignette:
with (self.expt_path /
'super_vignette.pickle').open('rb') as sv_file:
sv = pickle.load(sv_file)
calibrated_img = (img * calibration_img / ref_intensity *
self.BF_REF_INTENSITY).astype('uint16')
calibrated_img[~sv] = 0
return calibrated_img
else:
return (img * calibration_img / ref_intensity *
self.BF_REF_INTENSITY).astype('uint16')
def make_composite_maskfile_batch(data_path,
mask_path,
save_path,
data_str='',
mask_str=''):
data_fns = [
data_f for data_f in sorted(os.listdir(data_path))
if data_str in data_f
]
print('importing data images')
data_imgs = np.array([(freeimage.read(data_path + os.path.sep + data_f))
for data_f in data_fns])
print('importing mask images')
mask_imgs = np.array([
freeimage.read(mask_path + os.path.sep + mask_f) > 0
for data_f in data_fns for mask_f in sorted(os.listdir(mask_path))
if data_f[0:15] in mask_f if mask_str in mask_f
])
try:
os.stat(save_path)
except:
os.mkdir(save_path)
print('generating and saving composites')
comp = np.zeros(np.shape(data_imgs[[0]]))
print('got here')
for d_img, m_img, data_f in zip(data_imgs, mask_imgs, data_fns):
comp = colorize.scale(np.copy(d_img), output_max=254)
comp[m_img] = 255
#if data_f==data_fns[2]: return
freeimage.write(
comp.astype('uint8'),
save_path + os.path.sep + data_f[:-4] + 'composite' + data_f[-4:])
def computeFocusMeasures(temporal_radius=11, update_db=True, write_models=False, write_deltas=False, write_masks=False):
with sqlite3.connect(str(DPATH / "analysis/db.sqlite3")) as db:
db.row_factory = sqlite3.Row
non_vignette = freeimage.read(str(DPATH / "non-vignette.png")) != 0
vignette = non_vignette == 0
image_count = list(db.execute("select count() from images"))[0]["count()"]
positions = [row["well_idx"] for row in db.execute("select well_idx from wells where did_hatch")]
position_bgss = {position: WzBgs(2560, 2160, temporal_radius, non_vignette) for position in positions}
time_points = [row["name"] for row in db.execute("select name from time_points")]
image_idx = 0
for time_point in time_points:
print(time_point)
for position in positions:
print("", position)
acquisition_names = [
row["acquisition_name"]
for row in db.execute(
"select acquisition_name from images where well_idx=? and time_point=?", (position, time_point)
)
]
if acquisition_names:
bgs = position_bgss[position]
tasks = []
for acquisition_name in acquisition_names:
im_fpath = (
DPATH / "{:02}".format(position) / "{} {}_ffc.png".format(time_point, acquisition_name)
)
tasks.append(
pool.submit(
lambda an=acquisition_name, fn=_computeFocusMeasures, args=(
bgs,
im_fpath,
non_vignette,
update_db,
write_models,
write_deltas,
write_masks,
): (an, fn(*args))
)
)
for task in tasks:
acquisition_name, focus_measures = task.result()
if focus_measures is not None:
measure_names = sorted(focus_measures.keys())
q = "update images set " + ", ".join(
"{}=?".format(measure_name) for measure_name in measure_names
)
q += " where well_idx=? and time_point=? and acquisition_name=?"
v = [float(focus_measures[measure_name]) for measure_name in measure_names]
v.extend((position, time_point, acquisition_name))
list(db.execute(q, v))
image_idx += 1
print(" {:<10} {:%}".format(acquisition_name, image_idx / image_count))
try:
im = freeimage.read(str(DPATH / "{:02}".format(position) / "{} bf_ffc.png".format(time_point)))
bgs.updateModel(im)
except:
pass
db.commit()
def _computeSigmoidWeightedMeasures(db_lock, update_db, measure_antimask, x0, k, L):
weightedMaskedHighpassBrenner = getWeightedMaskedHighpassBrennerInstance((2560, 1600))
with db_lock, sqlite3.connect(str(DPATH / "analysis/db.sqlite3")) as db:
column_name = "x0:{},k:{}".format(x0, k)
time_point_well_idxs = list(
db.execute(
"select time_point, well_idx from ("
' select time_point, well_idx, sum(is_focused) as sif from images where acquisition_name!="bf" group by time_point, well_idx'
") where sif == 1"
)
)
for time_point, well_idx in time_point_well_idxs:
with db_lock, sqlite3.connect(str(DPATH / "analysis/db.sqlite3")) as db:
acquisition_names = [
row[0]
for row in db.execute(
"select acquisition_name from images where well_idx=? and time_point=?", (well_idx, time_point)
)
]
results = []
for acquisition_name in acquisition_names:
delta_fpath = (
DPATH
/ "{:02}".format(well_idx)
/ "{} {}_ffc wz_bgs_model_delta.tiff".format(time_point, acquisition_name)
)
if not delta_fpath.exists():
continue
image = freeimage.read(
str(DPATH / "{:02}".format(well_idx) / "{} {}_ffc.png".format(time_point, acquisition_name))
)
delta = freeimage.read(str(delta_fpath))
weights = logistic_sigmoid(delta, x0, k, L)
results.append(float(weightedMaskedHighpassBrenner.metric(image, measure_antimask, weights)))
if not results or not update_db:
continue
with db_lock, sqlite3.connect(str(DPATH / "analysis/db.sqlite3")) as db:
for result, acquisition_name in zip(results, acquisition_names):
image_id = list(
db.execute(
"select image_id from images where time_point=? and well_idx=? and acquisition_name=?",
(time_point, well_idx, acquisition_name),
)
)[0][0]
if list(db.execute("select count(*) from sigmoids where image_id=?", (image_id,)))[0][0] == 0:
list(
db.execute(
'insert into sigmoids (image_id, "{}") values (?, ?)'.format(column_name),
(image_id, result),
)
)
else:
list(
db.execute(
'update sigmoids set "{}"=? where image_id=?'.format(column_name), (result, image_id)
)
)
db.commit()
def select_sample(image_point, vignette_mask, total_sample_number, SVM_external_parameters):
'''
Given a mask in a_mask and an image file in image_file, sample points to use for texture classification according to selection_mode and total_sample_number.
'''
def random_select_from_mask(a_mask, sample_number, my_radius):
'''
Select sample_number pixels from a_mask which have a clear radius of length my_radius around themselves (i.e. they are not too close to the border of the image).
'''
# Ensure that the edges of the image are not included in the mask.
a_mask[:my_radius, :] = False
a_mask[-my_radius:, :] = False
a_mask[:, :my_radius] = False
a_mask[:, -my_radius:] = False
# Randomly select pixels from the mask.
pixel_number = a_mask[a_mask > 0].shape[0]
my_indices = np.ma.indices(my_image.shape)
selected_pixels = np.zeros(pixel_number).astype('bool')
selected_pixels[:int(total_sample_number/2)] = True
selected_pixels = np.random.permutation(selected_pixels)
selected_pixels = np.array([my_indices[0][a_mask > 0][selected_pixels], my_indices[1][a_mask > 0][selected_pixels]]).transpose()
return selected_pixels
# Rescale images and prepare some supporting variables.
worm_mask = freeimage.read(image_point + ' ' + 'hmask.png').astype('bool')
my_image = freeimage.read(image_point + ' ' + 'bf.png')
(my_image, worm_mask, vignette_mask) = scaled_images(my_image, worm_mask, vignette_mask, SVM_external_parameters)
my_radius = (SVM_external_parameters['square_size'] - 1)/2
# Select half pixels from the worm itself and half pixels from an area outside the worm but within 50 pixels of the worm.
selection_mode = SVM_external_parameters['sampling']
if selection_mode == 'Local_Worm':
distance_from_worm = scipy.ndimage.morphology.distance_transform_edt(np.invert(worm_mask)).astype('uint16')
background_mask = np.zeros(worm_mask.shape).astype('bool')
background_mask[distance_from_worm > 0] = True
background_mask[distance_from_worm > worm_mask.shape[0]//20] = False
background_mask[np.invert(vignette_mask)] = False
background_selected_pixels = random_select_from_mask(background_mask, int(total_sample_number/2), my_radius)
worm_selected_pixels = random_select_from_mask(worm_mask, int(total_sample_number/2), my_radius)
elif selection_mode == 'Hard_Negative':
crude_mask = freeimage.read(image_point + ' ' + 'mask.png').astype('bool')
negatives_mask = crude_mask & np.invert(worm_mask)
background_selected_pixels = random_select_from_mask(negatives_mask, int(total_sample_number/2), my_radius)
worm_selected_pixels = random_select_from_mask(worm_mask, int(total_sample_number/2), my_radius)
else:
raise BaseException('Invalid selection_mode passed from SVM_external_parameters[\'sampling\'].')
my_pixels = np.vstack([worm_selected_pixels, background_selected_pixels])
my_squares = [my_image[a_pixel[0] - my_radius: a_pixel[0] + 1 + my_radius, a_pixel[1] - my_radius: a_pixel[1] + 1 + my_radius] for a_pixel in my_pixels]
my_classifications = np.hstack([np.array([1]*int(total_sample_number/2)), np.array([0]*int(total_sample_number/2))])
my_features = feature_computer(my_squares, SVM_external_parameters)
my_samples = [my_features, my_classifications]
return my_samples
def __init__(self, rw, image_dir, out_dir):
"""
Parameters
rw - RisWidget object to load images into
image_dir - str/pathlib.Path to directory images for loading
out_dir - str/pathlib.Path to directory for saving overlays
"""
self.rw = rw
self.image_dir = pathlib.Path(image_dir)
self.out_dir = pathlib.Path(out_dir)
assert self.image_dir.exists()
assert self.out_dir.exists()
self.editing = False
# To be low-maintenance, just add everything to the flipbook layout
layout = Qt.QFormLayout()
widget = Qt.QWidget()
widget.setLayout(layout)
self.rw.flipbook.layout().addWidget(widget)
self.rw.add_painter()
self.rw.flipbook_pages.clear()
self.image_list = sorted(list(self.image_dir.glob('*.png')))
for image_path in self.image_list:
image = freeimage.read(image_path)
image_list = flipbook.ImageList()
image_list.append(rw_image.Image(data=image, name=image_path.name))
if (self.out_dir / image_path.name).exists():
image_list.append(
freeimage.read((self.out_dir / image_path.name)))
else:
image_shape = image.shape
if len(image_shape) == 3:
image_list.append(numpy.zeros_like(image))
else:
new_shape = list(image_shape) + [3]
image_list.append(
numpy.zeros(shape=new_shape).astype('uint8'))
rw.flipbook_pages.append(image_list)
self.rw.painter.brush_size.value = 9
self.clear = self._add_button(layout, 'Clear All',
self._on_clear_clicked)
self.reload = self._add_button(layout, 'Reload Overlay',
self._on_reload_clicked)
self.save = self._add_button(layout, 'Save Overlay',
self._on_save_clicked)
def read_corrected_bf(image_file, movement_key = ''):
'''
Read in an image at time_point and properly correct it for flatfield and metering.
'''
time_point = image_file.split(os.path.sep)[-1].split(' ')[0]
raw_image = freeimage.read(image_file)
flatfield_image = freeimage.read(os.path.sep.join(image_file.split(os.path.sep)[:-2]) + os.path.sep + 'calibrations' + os.path.sep + time_point + ' ' + 'bf_flatfield.tiff')
with open(os.path.sep.join(image_file.split(os.path.sep)[:-2]) + os.path.sep + 'experiment_metadata.json', 'r') as read_file:
metadata = json.loads(read_file.read())
time_reference = metadata['brightfield metering'][time_point]['ref_intensity']
corrected_image = raw_image*flatfield_image
corrected_image = corrected_image / time_reference * 11701.7207031
corrected_image = corrected_image.astype('uint16')
return corrected_image
def _makeMeasureInputSensitivityComparisonVizWorker(
measure_antimask,
measure_a,
measure_b,
bf_im_fpath,
model_im_fpath,
measure_a_im_fpath,
measure_a_idx_delta,
measure_a_delta_im_fpath,
measure_a_mask_im_fpath,
measure_b_im_fpath,
measure_b_idx_delta,
measure_b_delta_im_fpath,
measure_b_mask_im_fpath
):
r = _NS()
r.bf_im, r.bf_im_fpath = freeimage.read(str(bf_im_fpath)), bf_im_fpath
r.model_im = freeimage.read(str(model_im_fpath))
r.measure_a_im, r.measure_a_im_fpath = freeimage.read(str(measure_a_im_fpath)), measure_a_im_fpath
r.measure_a_idx_delta = measure_a_idx_delta
r.measure_a_delta_im = freeimage.read(str(measure_a_delta_im_fpath))
r.measure_a_mask_im = freeimage.read(str(measure_a_mask_im_fpath))
r.measure_a_transformed_im = _apply_measure_transform(measure_a, r.measure_a_im, measure_antimask, r.measure_a_delta_im, r.measure_a_mask_im)
r.measure_a_transformed_im_b = _apply_measure_transform(measure_b, r.measure_a_im, measure_antimask, r.measure_a_delta_im, r.measure_a_mask_im)
r.measure_b_im, r.measure_b_im_fpath = freeimage.read(str(measure_b_im_fpath)), measure_b_im_fpath
r.measure_b_idx_delta = measure_b_idx_delta
r.measure_b_delta_im = freeimage.read(str(measure_b_delta_im_fpath))
r.measure_b_mask_im = freeimage.read(str(measure_b_mask_im_fpath))
r.measure_b_transformed_im = _apply_measure_transform(measure_b, r.measure_b_im, measure_antimask, r.measure_b_delta_im, r.measure_b_mask_im)
r.measure_b_transformed_im_a = _apply_measure_transform(measure_a, r.measure_b_im, measure_antimask, r.measure_b_delta_im, r.measure_b_mask_im)
return r
def train_texture_SVM_sample(data_points, regression_variable,
samples_per_image, texture_model):
'''
Get sample histograms for texture classification SVM training.
'''
histogram_array = []
age_array = []
for a_point in data_points:
do_point = True
subdir = os.path.sep.join(a_point.split(os.path.sep)[:-1])
with open(subdir + os.path.sep + 'position_metadata_extended.json',
'r') as read_file:
my_metadata = json.loads(read_file.read())
metadata_dict = {
my_metadata[i]['timepoint']:
my_metadata[i][regression_variable.lower()]
for i in range(0, len(my_metadata))
}
egg_age_dict = {
my_metadata[i]['timepoint']: my_metadata[i]['egg_age']
for i in range(0, len(my_metadata))
}
# Exclude larval animals from ghost_age computation.
if regression_variable.lower() == 'ghost_age':
time_name = a_point.split(os.path.sep)[-1].split(' ')[0]
if egg_age_dict[time_name] > 0:
do_point = True
else:
do_point = False
if do_point:
my_time = a_point.split(os.path.sep)[-1].split(' ')[0]
my_age = metadata_dict[my_time] / 24
mask_file = freeimage.read(a_point)
image_file = freeimage.read(a_point.replace('hmask.png', 'bf.png'))
my_samples = sample_8star(image_file,
mask_file,
17,
sample_number=samples_per_image)
my_codebook = np.zeros(texture_model.cluster_vectors.shape[0])
for a_sample in my_samples:
my_codebook[texture_model.classify_vector(a_sample)] += 1
my_codebook = my_codebook / np.sum(my_codebook)
histogram_array.append(my_codebook)
age_array.append(my_age)
return (histogram_array, age_array)
def preprocess_image(self, i):
downscale = self.downscale
lab_frame_image = freeimage.read(self.timepoint_list[i].image_path('bf'))
lab_frame_image = lab_frame_image.astype(numpy.float32)
height, width = lab_frame_image.shape[:2]
try:
metadata = self.timepoint_list[i].position.experiment.metadata
optocoupler = metadata['optocoupler']
except KeyError:
optocoupler = 1
mode = process_images.get_image_mode(lab_frame_image, optocoupler=optocoupler)
#### DownSample the image
if downscale > 0 and downscale != 1:#and set_name!='train':
#t_size = (int(width / downscale), int(height / downscale))
shrink_image = pyramid.pyr_down(lab_frame_image, downscale=downscale)
#shrink_image = numpy.clip(shrink_image, 0, 40000)
else:
shrink_image = lab_frame_image
shrink_image = shrink_image.astype(numpy.float32)
## scale the image pixel value into a trainable range
# map image image intensities in range (100, 2*mode) to range (0, 2)
bf = colorize.scale(shrink_image, min=100, max=2*mode, output_max=2)
# now shift range to (-1, 1)
bf -= 1
return bf
def set_well(self, index):
self.well_index = index
well_name = self.well_names[index]
images = [freeimage.read(str(image)) for image in sorted(self.date_dir.glob('well_images/{}-*.png'.format(well_name)))]
self.animator.start(images)
self.well = self.well_names[index]
self.set_status(self.statuses[self.well_index])
def generate_grader(generate_specific_worm_from_parameters, mask_worm, mask_scorer, my_PCA, standard_mask_file, which_fixed = None, fixed_parameters = None): ''' Generates a function that will grade a worm generated from parameters against standard_mask. It takes as input: generate_specific_worm_from_parameters: a function which will generate a set of outline points when given the mean worm as a set of outline points, the PCA object with the linear weights for each PC, a point around which to center the worm, and a set of weights to apply to the PC object. mask_worm: a function that will take a worm from a set of points and make a rasterized mask for comparison. mask_scorer: a function that will score the "goodness" of fit when given two masks of compatible size. my_PCA: an object with results from principal component analysis of the data. standard_mask_file: the file containing the mask that we want to fit our worm to. ''' standard_mask = freeimage.read(standard_mask_file) def my_grader(my_parameters): if which_fixed == None: my_fixed_parameters = np.array([False]*my_parameters.shape[0]) my_which_fixed = np.array([False]*my_parameters.shape[0]) else: my_fixed_parameters = fixed_parameters my_which_fixed = which_fixed my_real_parameters = np.zeros(my_which_fixed.shape) my_real_parameters[my_which_fixed] = my_fixed_parameters[my_which_fixed] my_real_parameters[np.invert(my_which_fixed)] = my_parameters test_worm = generate_specific_worm_from_parameters(my_PCA, my_real_parameters) test_mask = mask_worm(test_worm, my_PCA, my_real_parameters, standard_mask.shape) my_score = mask_scorer(test_mask, standard_mask) return my_score return my_grader
def _start_autofocus(self,
metric='brenner',
metric_kws=None,
metric_mask=None,
metric_filter_period_range=None):
if isinstance(metric_mask, str):
metric_mask = freeimage.read(metric_mask) > 0
shape = self._camera.get_aoi_shape()
if isinstance(metric, str):
if metric in self._METRICS:
metric = self._METRICS[metric]
elif ':' in metric:
path, metric = metric.split(':')
metric = runpy.run_path(path)[metric]
else:
raise ValueError(
'"metric" must be the name of a known metric or formatted as "/path/to/file.py:function"'
)
if metric_kws is None:
metric_kws = {}
# check if metric is a class at all before asking if it's our subclass of interest:
if isinstance(metric, type) and issubclass(metric,
AutofocusMetricBase):
return metric(shape,
mask=metric_mask,
fft_period_range=metric_filter_period_range,
**metric_kws)
else:
assert callable(metric)
return AutofocusMetric(metric,
shape,
mask=metric_mask,
fft_period_range=metric_filter_period_range,
**metric_kws)
def generate_images_from_files(image_files,
upper_left=(0, 0),
lower_right=(None, None),
min=None,
max=None,
gamma=1,
color=(255, 255, 255)):
"""Load images from a file, optionally cropping, scaling, and colorizing them
before converting to 8 bit. Each image is yielded as it is loaded and transformed.
Parameters:
image_files: list of image files to read
upper_left: (x, y) coordinates of upper left corner of region to include
lower_right: (x, y) coordinates of lower right corner of region to include
(use None to use full image extent).
min, max: image intensity values to map to black and white, respectively.
If None, use each image's min and max value.
gamma: gamma value for intensity transformation.
color: 8-bit (R,G,B) color to map "white" to. By default this is white:
(255, 255, 255). To map the brightest possible color to green instead
use (0, 255, 0), for example.
"""
x1, y1 = upper_left
x2, y2 = lower_right
for image_file in image_files:
image = freeimage.read(image_file)
cropped_image = image[x1:x2, y1:y2]
scaled_image = colorize.scale(cropped_image,
min,
max,
gamma,
output_max=1)
colorized_8bit_image = colorize.color_tint(scaled_image,
color).astype(numpy.uint8)
yield colorized_8bit_image
def process_centerline_dir(source_dir, microns_per_pixel):
source_dir = pathlib.Path(source_dir)
out_dir = source_dir / 'individual_centerlines'
out_dir.mkdir(exist_ok=True)
centerline_data = {}
centerline_data_entries = ['name', 'length']
mask_data = {}
[mask_data.setdefault(entry,[]) for entry in centerline_data_entries]
for centerline_image_path in sorted(source_dir.iterdir()):
if centerline_image_path.suffix[1:] != 'png':
continue
aggregate_centerline_image = freeimage.read(centerline_image_path)
masks = parse_aggregate_centerlines(aggregate_centerline_image)
for mask_num, mask in enumerate(masks):
mask_name = centerline_image_path.stem + f'_{mask_num}'
print(mask_name)
freeimage.write(mask.astype('uint8')*255, out_dir / (mask_name+'.png'))
center_tck, _ = worm_spline.pose_from_mask(mask) # Toss the widths
try:
length = spline_geometry.arc_length(center_tck) * microns_per_pixel
mask_data['name'].append(mask_name)
mask_data['length'].append(length)
except TypeError:
print(f'Warning: couldn\'t find centerline for {mask_name} (px location {list(numpy.where(mask))})')
with (out_dir / 'measurements.txt').open('w+') as measurement_file:
measurement_file.write('\t'.join(centerline_data_entries)+'\n')
for data in zip(*[mask_data[entry] for entry in centerline_data_entries]):
measurement_file.write('\t'.join([str(item) for item in data])+'\n')
def warp_image(spine_tck, width_tck, image_file, warp_file):
"""Warp an image of a worm to a specified place
Parameters:
spine_tck: parametric spline tck tuple corresponding to the centerline of the worm
width_tck: non-parametric spline tck tuple corresponding to the widths of the worm
image_file: path to the image to warp the worm from
warp_file: path where the warped worm image should be saved to
"""
image = freeimage.read(image_file)
warp_file = pathlib.Path(warp_file)
#730 was determined for the number of image samples to take perpendicular to the spine
#from average length of a worm (as determined from previous tests)
warped = resample.warp_image_to_standard_width(image, spine_tck, width_tck,
width_tck,
int(tck[0][-1] // 5))
#warped = resample.sample_image_along_spline(image, spine_tck, 730)
mask = resample.make_mask_for_sampled_spline(warped.shape[0],
warped.shape[1], width_tck)
warped = colorize.scale(warped).astype('uint8')
warped[~mask] = 255
print("writing warped worm to :" + str(warp_file))
#return warped
if not warp_file.exists():
warp_file.parent.mkdir(exist_ok=True)
freeimage.write(
warped, warp_file
) # freeimage convention: image.shape = (W, H). So take transpose.
def _read_page_task(self, task_page):
task_page.ims = [
freeimage.read(str(image_fpath))
for image_fpath in task_page.im_fpaths
]
Qt.QApplication.instance().postEvent(self,
_ReadPageTaskDoneEvent(task_page))
def color_dots(image_paths, save_paths, color_dot_location_lists):
'''
Draws colored dots on images from image_paths according to color_dot_location_lists, and saves them out to save_paths.
'''
color_dot_location_lists = np.round(color_dot_location_lists)
for i in range(0, len(image_paths)):
image_array = freeimage.read(image_paths[i])
print('Drawing dots for ' + image_paths[i] + '.')
if len(image_array.shape) == 2:
my_width, my_height = image_array.shape
color_array = np.empty((my_width, my_height, 3), dtype = image_array.dtype)
color_array[:, :, 0] = image_array.copy()
color_array[:, :, 1] = image_array.copy()
color_array[:, :, 2] = image_array.copy()
elif len(image_array.shape) == 3:
color_array = image_array
color_array[color_dot_location_lists[0][i][0], color_dot_location_lists[0][i][1], :] = [0, 0, 0]
if len(color_dot_location_lists) > 1:
color_array[color_dot_location_lists[1][i][0], color_dot_location_lists[1][i][1], :] = [0, 0, 0]
if len(color_dot_location_lists) > 2:
color_array[color_dot_location_lists[2][i][0], color_dot_location_lists[2][i][1], :] = [0, 0, 0]
color_array[color_dot_location_lists[0][i][0], color_dot_location_lists[0][i][1], 0] = -1
if len(color_dot_location_lists) > 1:
color_array[color_dot_location_lists[1][i][0], color_dot_location_lists[1][i][1], 1] = -1
if len(color_dot_location_lists) > 2:
color_array[color_dot_location_lists[2][i][0], color_dot_location_lists[2][i][1], 2] = -1
freeimage.write(color_array, save_paths[i])
return
def assemble_image_sequence(out_dir, well_name, crop=True):
out_dir = pathlib.Path(out_dir)
well_images = collections.defaultdict(list)
for img in sorted(out_dir.glob('*/well_images/{}-*.png'.format(well_name))):
date = img.parent.parent.name
well_images[date].append(freeimage.read(str(img)))
dates, images = zip(*sorted(well_images.items()))
maxlen = max(len(ims) for ims in images)
shapes = numpy.array([ims[0].shape for ims in images])
if crop:
shape = shapes.min(axis=0)
else:
shape = shapes.max(axis=0)
new_images = []
for ims in images:
oldshape = ims[0].shape
extra = numpy.abs(shape - oldshape)/2
if numpy.any(extra):
extra = list(zip(numpy.ceil(extra).astype(int), extra.astype(int)))
if crop:
(xl, xh), (yl, yh) = extra
xh = -xh if xh else None
yh = -yh if yh else None
out = [i[xl:xh, yl:yh] for i in ims]
else:
out = [numpy.pad(i, extra, 'constant') for i in ims]
else:
out = ims
extra_ims = maxlen - len(ims)
if extra_ims:
out += out[-1:]*extra_ims
new_images.append(out)
images_out = [numpy.concatenate([ims[i] for ims in new_images]) for i in range(maxlen)]
return images_out, shape
def _find_worms_task(image_path, well):
print('Finding worm '+well)
image = freeimage.read(image_path)
well_mask, edges, worm_mask = find_worm.find_worm_from_brightfield(image)
freeimage.write(well_mask.astype(numpy.uint8)*255, image_path.parent / (well + '_well_mask.png'))
freeimage.write(worm_mask.astype(numpy.uint8)*255, image_path.parent / (well + '_worm_mask.png'))
return is_valid_mask(worm_mask)
def straighten_worms_from_rw(rw, warp_path):
warp_dir = pathlib.Path(warp_path)
if not warp_dir.exists():
warp_dir.mkdir()
current_idx = rw.flipbook.current_page_idx
img, sx, sy = rw.flipbook.pages[current_idx].img_path
bf_img_file = get_bf_image(img)
bf_img = freeimage.read(bf_img_file)
#need to crop image like the risWidget ones
x = slice(max(0, sx.start - 50), min(bf_img.shape[0], sx.stop + 50))
y = slice(max(0, sy.start - 50), min(bf_img.shape[1], sy.stop + 50))
crop_img = bf_img[x, y]
warp_name = bf_img_file.stem.split(" ")[0] + "_" + str(
current_idx) + "_warp.png"
traceback = rw.flipbook.pages[current_idx].spline_data
dist = rw.flipbook.pages[current_idx].dist_data
#print("worm length: ",len(traceback))
tck = skeleton.generate_splines(traceback, dist)
#print("tck length: ",len(tck))
width_tck = skeleton.width_spline(traceback, dist)
warp_image(tck, width_tck, crop_img, warp_dir.joinpath(warp_name))
def flatfield_correct(im_fpath, ff_fpath, ffc_fpath):
if not im_fpath.exists():
return False, 'skipping "{}" (file not found)'.format(str(im_fpath))
if not ff_fpath.exists():
return False, 'skipping "{}" (flatfield reference image file "{}" not found)'.format(str(ff_fpath))
try:
im = freeimage.read(str(im_fpath))
ff = freeimage.read(str(ff_fpath))
ffc = im.astype(numpy.float32) * ff
ffc *= (65535.0 * 0.9) / float(numpy.percentile(ffc, 98))
ffc[ffc < 0] = 0
ffc[ffc > 65535] = 65535
freeimage.write(ffc.astype(numpy.uint16), str(ffc_fpath), freeimage.IO_FLAGS.PNG_Z_BEST_SPEED)
except Exception as e:
return False, 'exception while correcting "{}": {}'.format(str(im_fpath), e)
return True, '{} done'.format(str(im_fpath))
def corrected_worm_frame_image(position_root, timepoint, image_type, center_tck, optocoupler=None):
if image_type == 'bf':
image = freeimage.read(position_root / f'{timepoint} {image_type}.png')
image = pin_image_mode(image, optocoupler=optocoupler)
else:
image = flatfield_correct(position_root, timepoint, image_type)
return worm_spline.to_worm_frame(image, center_tck)
def lawn_maker(bf_files, super_vignette):
'''
Find the bacterial lawn in one worm image.
'''
lawn_masks = []
for a_bf_file in bf_files:
# Prepare a worm image for use in lawn-finding.
renormalized_image = freeimage.read(a_bf_file)
renormalized_image = cv2.medianBlur(renormalized_image, 3)
renormalized_image = imageOperations.renormalize_image(renormalized_image)
# Remove extraneous edges and out-of-lawn junk by finding the lawn and also applying an "ultra-vignette" mask.
ultra_vignette = scipy.ndimage.morphology.binary_erosion(super_vignette, iterations = 10)
my_edges = skimage.feature.canny(renormalized_image, sigma = 0.02)
my_edges[np.invert(ultra_vignette)] = False
my_edges = scipy.ndimage.morphology.binary_dilation(my_edges, iterations = 10)
my_lawn = scipy.ndimage.morphology.binary_fill_holes(my_edges)
try:
my_lawn = zplib_image_mask.get_largest_object(my_lawn).astype('bool')
my_lawn = scipy.ndimage.morphology.binary_erosion(my_lawn, iterations = 10)
my_lawn = zplib_image_mask.get_largest_object(my_lawn).astype('bool')
except:
my_lawn = np.zeros(my_lawn.shape).astype('bool')
lawn_masks.append(my_lawn)
my_lawn = np.max(np.array(lawn_masks), axis = 0)
return my_lawn
def train_texture_samples(subdir, samples_dict, regression_variable, endings,
samples_per_image, age_ranges):
'''
Read in samples for texture classification.
'''
# Organize some information.
total_files = os.listdir(subdir)
data_points = [
' '.join(a_file.split('.')[0].split(' ')[0:-1])
for a_file in total_files if a_file.split(' ')[-1] == 'hmask.png'
]
with open(subdir + os.path.sep + 'position_metadata_extended.json',
'r') as read_file:
my_metadata = json.loads(read_file.read())
metadata_dict = {
my_metadata[i]['timepoint']:
my_metadata[i][regression_variable.lower()]
for i in range(0, len(my_metadata))
}
egg_age_dict = {
my_metadata[i]['timepoint']: my_metadata[i]['egg_age']
for i in range(0, len(my_metadata))
}
# Exclude larval animals from ghost_age computation.
if regression_variable.lower() == 'ghost_age':
data_points = [
a_point for a_point in data_points if egg_age_dict[a_point] > 0
]
# Actualyl sample my images.
for a_point in data_points:
my_age = metadata_dict[a_point] / 24
if regression_variable.lower() == 'ghost_age':
my_age = abs(my_age)
mask_file = freeimage.read(subdir + os.path.sep + a_point + ' ' +
endings[1])
image_file = freeimage.read(subdir + os.path.sep + a_point + ' ' +
endings[0])
my_samples = sample_8star(image_file,
mask_file,
17,
sample_number=samples_per_image)
for age_range in age_ranges:
if age_range[0] <= my_age < age_range[1]:
samples_dict[age_range].extend(my_samples)
return samples_dict
def normalized_bf_image(self, i):
bf = freeimage.read(self.timepoint_list[i].image_path('bf'))
mode = process_images.get_image_mode(bf, optocoupler=self.timepoint_list.optocoupler(i))
# map image image intensities in range (100, 2*mode) to range (0, 2)
bf = colorize.scale(bf, min=100, max=2*mode, output_max=2)
# now shift range to (-1, 1)
bf -= 1
return bf
def load_images_fromfns(image_fns, filter_kw='', master_vig=None):
image_list = []
for image_fn in image_fns:
image = freeimage.read(image_fn)
if master_vig is not None:
image[master_vig == 0] = 0
image_list.append(image)
return image_list
def OLDoverallBackgroundSubtract(data_dir, match_string, temporal_radius, save_dir):
'''
Do background subtraction to find worms.
'''
my_files = sorted(os.listdir(data_dir))
my_files = [a_file for a_file in my_files if match_string == a_file.split('_')[-1]]
# Intialize my special background context.
temp_folder = save_dir + '\\' + 'temp'
try:
os.stat(temp_folder)
except:
os.mkdir(temp_folder)
for i in range(0, temporal_radius):
shutil.copy(data_dir + '\\' + my_files[i], temp_folder + '\\' + my_files[i])
# Run the actual simple subtraction, saving out masked files.
for i in range(temporal_radius, len(my_files)-temporal_radius):
#context_files = [freeimage.read(data_dir + '\\' + my_files[j]) for j in range(i-temporal_radius, i+temporal_radius+1)]
context_files = [freeimage.read(data_dir + '\\' + my_files[j]) for j in range(i-temporal_radius, i+1)]
raw_file = freeimage.read(data_dir + '\\' + my_files[i])
(simple_foreground_file, background_file) = simple_running_median_subtraction(raw_file, context_files)
thresholded_mask = percentile_floor(simple_foreground_file, threshold_proportion = 0.975)
final_mask = clean_dust_and_holes(thresholded_mask)
raw_file[final_mask.astype('bool')] = background_file[final_mask.astype('bool')]
freeimage.write(raw_file, temp_folder + '\\' + my_files[i])
# Fill in remaining tail files.
for i in range(len(my_files)-temporal_radius, len(my_files)):
shutil.copy(data_dir + '\\' + my_files[i], temp_folder + '\\' + my_files[i])
# Now let's do it for real!
for i in range(temporal_radius, len(my_files)-temporal_radius):
context_files = [freeimage.read(temp_folder + '\\' + my_files[j]) for j in range(i-temporal_radius, i+temporal_radius+1)]
raw_file = freeimage.read(data_dir + '\\' + my_files[i])
(simple_foreground_file, background_file) = simple_running_median_subtraction(raw_file, context_files)
thresholded_pic = percentile_floor(simple_foreground_file, threshold_proportion = 0.975)
final_mask = clean_dust_and_holes(thresholded_pic)
freeimage.write(final_mask, save_dir + '\\' + my_files[i])
return
def overlay_masks(rw, position_directory):
position_directory = pathlib.Path(position_directory)
expt_dir = position_directory.parent
position_annotations = load_data.read_annotations(expt_dir)[position_directory.name]
files_to_load = []
page_names = []
global_positions, timepoint_annotations = position_annotations
for timepoint, timepoint_data in timepoint_annotations.items():
image_key = position_directory.name + '_' + timepoint
image = freeimage.read(str(position_directory / (timepoint + ' bf.png')))
mask_file = expt_dir / 'derived_data' / 'mask' / position_directory.name / (timepoint + ' bf.png')
if mask_file.exists():
mask_image = freeimage.read(str(expt_dir / 'derived_data' / 'mask' / position_directory.name / (timepoint + ' bf.png'))) > 0
files_to_load.append([image, mask_image])
else:
files_to_load.append([image])
rw.flipbook_pages = files_to_load
def save_mask_lcc(experiment_root):
experiment_root = pathlib.Path(experiment_root)
mask_root = experiment_root / 'derived_data' / 'mask'
for position_mask_root in sorted(mask_root.iterdir()):
for mask_file in sorted(position_mask_root.iterdir()):
mask_image = freeimage.read(str(mask_file)) > 0
new_mask = mask.get_largest_object(mask_image).astype(numpy.uint8)
freeimage.write(new_mask*255, str(mask_file))
def _start_autofocus(self, focus_filter_period_range, focus_filter_mask,
**camera_state):
camera_state.update(self._CAMERA_MODE)
self._camera.push_state(**camera_state)
if focus_filter_mask is not None:
focus_filter_mask = freeimage.read(focus_filter_mask) > 0
self._metric = BrennerAutofocus(self._camera.get_aoi_shape(),
focus_filter_period_range,
focus_filter_mask)
def size_histogram(a_directory):
'''
return a list of sizes
'''
my_files = [a_directory + os.path.sep + a_file for a_file in os.listdir(a_directory)]
my_images = [freeimage.read(a_file) for a_file in my_files]
my_masks = [an_image[:, :, 0].astype('bool') for an_image in my_images]
my_sizes = [np.sum(a_mask) for a_mask in my_masks]
return my_sizes
def crop_img(rw):
current_idx = rw.flipbook.current_page_idx
img, sx, sy = rw.flipbook.pages[current_idx].img_path
bf_img_file = get_bf_image(img)
bf_img = freeimage.read(bf_img_file)
#need to crop image like the risWidget ones
x = slice(max(0, sx.start - 50), min(bf_img.shape[0], sx.stop + 50))
y = slice(max(0, sy.start - 50), min(bf_img.shape[1], sy.stop + 50))
crop_img = bf_img[x, y]
def overallBackgroundSubtract(data_dpath, match_glob, temporal_radius, save_dpath, save_dpath2 = '', save_dpath3 = '', demonstration_mode = False):
'''
Do background subtraction to find worms. This uses only past data, masking out the worms to create a background that won't disappear once the worm stops moving.
'''
data_dpath = Path(data_dpath)
save_dpath = Path(save_dpath)
if save_dpath2:
save_dpath2 = Path(save_dpath2)
if save_dpath3:
save_dpath3 = Path(save_dpath3)
my_file_fpaths = sorted(data_dpath.glob(match_glob))
# Initialize my special background context.
temp_dpath = save_dpath / 'temp'
if not temp_dpath.exists():
temp_dpath.mkdir(parents=True)
for i in range(0, temporal_radius):
shutil.copy(str(my_file_fpaths[i]), str(temp_dpath / my_file_fpaths[i].name))
# Run the actual simple subtraction, saving out masked files.
context_files = [freeimage.read(str(my_file_fpaths[j])) for j in range(0, temporal_radius)]
for i in range(temporal_radius, len(my_file_fpaths)):
real_raw_file = freeimage.read(str(my_file_fpaths[i]))
raw_file = real_raw_file.copy()
context_files.append(raw_file)
(foreground_file, background_file) = simple_running_median_subtraction(raw_file, context_files)
thresholded_mask = percentile_floor(foreground_file, threshold_proportion = 0.975)
final_mask = clean_dust_and_holes(thresholded_mask)
raw_file[final_mask.astype('bool')] = background_file[final_mask.astype('bool')]
if demonstration_mode:
freeimage.write(real_raw_file, str(save_dpath / my_file_fpaths[i].name))
freeimage.write(background_file, str(save_dpath2 / my_file_fpaths[i].name))
freeimage.write(final_mask, str(save_dpath3 / my_file_fpaths[i].name))
if not demonstration_mode:
freeimage.write(raw_file, str(temp_dpath / my_file_fpaths[i].name))
freeimage.write(final_mask,str(save_dpath / my_file_fpaths[i].name))
context_files = context_files[1:]
return
def view_slices_experiment(expt_dir,
positions,
calibration_dir,
super_vignette,
rw,
measurement_name=None,
slices=None):
"""Straighten a bunch of worms and then put them up on risWidget for QC
"""
expt_dir = pathlib.Path(expt_dir)
annotation_dir = expt_dir / 'annotations'
calibration_dir = pathlib.Path(calibration_dir)
super_vignette = pickle.load(open(super_vignette, 'rb'))
worm_measurements = collections.OrderedDict()
#iterate through the worms
for worm, timepoints in positions.items():
print("Measuring worm: " + worm)
annotation_file = list(annotation_dir.glob('*' + worm + '.pickle'))[0]
_, annotations = pickle.load(open(annotation_file, 'rb'))
for tp, images in timepoints.items():
#normalize gfp image
raw_image = freeimage.read(images[0])
bf_image = freeimage.read(images[1])
flatfield_image = freeimage.read(calibration_dir /
(tp + " fl_flatfield.tiff"))
corrected_gfp = normalize_gfp_image(raw_image, super_vignette,
flatfield_image)
#warp worm to unit worm
tcks = annotations[tp]['pose']
warped_image, mask = warp_image(tcks, corrected_gfp)
warped_bf, mask = warp_image(tcks, bf_image)
if slices is not None:
for measure, slc in slices.items():
#print("Measuring slice: "+measure)
image_slice, mask_slice = slice_worms(
warped_image, mask, slc)
bf_slice, mask_slice = slice_worms(warped_bf, mask, slc)
#print(image_slice, bf_slice)
rw.flipbook_pages.append(
[image_slice, bf_slice, mask_slice])
rw.flipbook_pages[-1].name = (worm + " " + tp + " " +
measure)
def normalized_bf_image(timepoint):
"""Given a timepoint, return a normalized brightfield image."""
bf = freeimage.read(timepoint.image_path('bf'))
mode = process_images.get_image_mode(
bf, optocoupler=timepoint.position.experiment.metadata['optocoupler'])
# map image image intensities in range (100, 2*mode) to range (0, 2)
bf = colorize.scale(bf, min=100, max=2 * mode, output_max=2)
# now shift range to (-1, 1)
bf -= 1
return bf
def calculate_iou(prediction, ground_truth, plot_val=False, save_val=False, save_dir=None, pyr=False):
"""Find intersection over union for 2 images
Prediction and ground_truth are boolean arrays
"""
pred_orig = freeimage.read(prediction)
gt_orig = freeimage.read(ground_truth)
folder, worm_id = ground_truth.parts[-2:]
worm_id = worm_id.split(' ')[0]
#get rid of stuff touching the edges
if pyr:
pred = mask.remove_edge_objects(pred_orig)
gt = mask.remove_edge_objects(gt_orig)
#find the worm in the mask (largest object presumably)
pred_image = mask.get_largest_object(pred)
ground_truth = mask.get_largest_object(gt
)
else:
#find the worm in the mask (largest object presumably)
pred_image = mask.get_largest_object(pred_orig)
ground_truth = mask.get_largest_object(gt_orig)
#find intersect and union
intersect = (pred_image & ground_truth)
union = (pred_image|ground_truth)
#plot for validation
if save_val:
#maybe change this to not be this
save_file = save_dir+"/"+folder+"_"+worm_id+"_val.png"
#if willie:
# save_file = '/home/nicolette/Documents/lab_stuff/worm_segmentation_validation/Willie_iou/'+folder+"_"+worm_id+"_val.png"
#else:
# save_file = '/home/nicolette/Documents/lab_stuff/MATLAB_scripts/binarySeg_v2/figFolder/background_sub_pyr/iou/'+prediction.stem+"_val.png"
print("saving: "+save_file)
plot_validation(pred_orig, gt_orig, intersect, union, save_val=True, save_file=save_file)
if plot_val:
plot_validation(pred_orig, gt_orig, intersect, union)
return intersect.sum()/union.sum()
def extract_mask_fromcomposite_batch(comp_path,
save_path,
comp_str='',
save_str=''):
try:
os.stat(save_path)
except:
os.mkdir(save_path)
[freeimage.write(
(freeimage.read(comp_path+os.path.sep+comp_f) == 255).astype('uint16')*65535,save_path+os.path.sep+comp_f[:-4]+save_str+comp_f[-4:]) \
for comp_f in sorted(os.listdir(comp_path))]
def read_grayscale_array_from_image_file(filename, warn = True):
"""Read an image from disk into a 2-D grayscale array, converting from color if necessary.
If 'warn' is True, issue a warning when arrays are converted from color to grayscale.
"""
image_array = freeimage.read(filename)
if len(image_array.shape) == 3:
image_array = make_grayscale_array(image_array)
if warn:
warn_tools.warn('Image %s converted from RGB to grayscale: intensity values have been scaled and combined.'%filename)
return image_array
def get_worm(directory_bolus, a_worm, a_time, box_size = None, get_mode = 'worm'):
'''
Get a cut-out of a_worm at a_time.
'''
# Cut out the chunk that I want from the right image.
image_file = directory_bolus.working_directory + os.path.sep + a_worm + os.path.sep + a_time + ' ' + 'bf.png'
if get_mode == 'fluorescence':
worm_directory = [a_dir for a_dir in directory_bolus.data_directories if ' '.join(a_worm.split(' ')[:-2]) + ' Run ' + a_worm.split(' ')[-2] in a_dir][0] + os.path.sep + a_worm.split(' ')[-1]
image_file = worm_directory + os.path.sep + a_time + ' ' + 'green_yellow_excitation_autofluorescence.png'
my_image = corrected_fluorescence(image_file)
else:
my_image = freeimage.read(image_file)
# Get the right mask.
if get_mode in ['worm', 'fluorescence']:
mask_file = directory_bolus.working_directory + os.path.sep + a_worm + os.path.sep + a_time + ' ' + 'mask.png'
elif get_mode == 'lawn':
worm_directory = [a_dir for a_dir in directory_bolus.data_directories if ' '.join(a_worm.split(' ')[:-2]) + ' Run ' + a_worm.split(' ')[-2] in a_dir][0] + os.path.sep + a_worm.split(' ')[-1]
mask_file = worm_directory + os.path.sep + 'great_lawn.png'
if os.path.exists(mask_file):
my_mask = freeimage.read(mask_file).astype('bool')
else:
raise BaseException('Can\'t access files.')
# Mask out non-worm for fluorescence mode.
if get_mode == 'fluorescence':
my_image[np.invert(my_mask)] = 0
# Ensure that the final square is the right shape.
(square_mask, actual_size) = bound_box(my_mask, box_size = box_size)
my_square = my_image[square_mask]
my_square = np.reshape(my_square, actual_size)
box_size = (box_size, box_size)
if actual_size[0] < box_size[0]:
my_square = np.concatenate([my_square, np.zeros((box_size[0] - actual_size[0], my_square.shape[1])).astype('uint16')], axis = 0)
if actual_size[1] < box_size[1]:
my_square = np.concatenate([my_square, np.zeros((my_square.shape[0], box_size[1] - actual_size[1])).astype('uint16')], axis = 1)
if (my_square.shape[0]//2, my_square.shape[1]//2) != box_size:
raise BaseException('')
return my_square
def lawn_distribution(lawn_folder):
'''
Find the distribution of lawn sizes in a folder.
'''
def size_object(a_mask):
'''
Find the two points farthest apart in a mask and return their distance from each other.
'''
locations = np.array(np.ma.nonzero(a_mask)).transpose()
random_point = locations[0]
point_a = locations[np.linalg.norm(locations - random_point, axis = 1).argmax()]
point_b = locations[np.linalg.norm(locations - point_a, axis = 1).argmax()]
my_distance = np.linalg.norm(point_a - point_b)
return my_distance
my_lawns = [freeimage.read(lawn_folder + os.path.sep + an_image).astype('bool') for an_image in os.listdir(lawn_folder) if an_image != 'vignette_mask.png']
my_lawns = [zplib_image_mask.get_largest_object(my_lawn).astype('bool') for my_lawn in my_lawns]
my_vignette = freeimage.read(lawn_folder + os.path.sep + 'vignette_mask.png').astype('bool')
lawn_sizes = [size_object(my_lawn) for my_lawn in my_lawns]
field_size = size_object(my_vignette)
return (lawn_sizes, field_size)
def egg_outlines(a_directory):
'''
For each outlined egg mask in a_directory, convert it to a grayscale 8-bit image with the outline in white and the rest of the image in black.
'''
for a_subdir in os.listdir(a_directory):
for an_image in os.listdir(a_directory + os.path.sep + a_subdir):
if an_image.split(' ')[-1] == 'bf.png':
filepath = a_directory + os.path.sep + a_subdir + os.path.sep + an_image
my_image = freeimage.read(filepath)
masked_conversion = fill_colored_outline(my_image, egg_mode = True)
outline_conversion = fill_colored_outline(my_image, egg_mode = True, outline_only = True)
freeimage.write(masked_conversion, filepath.replace('bf', 'emask'))
freeimage.write(outline_conversion, filepath.replace('bf', 'outline'))
return
def train_texture_SVM_sample(data_points, regression_variable, samples_per_image, texture_model):
'''
Get sample histograms for texture classification SVM training.
'''
histogram_array = []
age_array = []
for a_point in data_points:
do_point = True
subdir = os.path.sep.join(a_point.split(os.path.sep)[:-1])
with open(subdir + os.path.sep + 'position_metadata_extended.json', 'r') as read_file:
my_metadata = json.loads(read_file.read())
metadata_dict = {my_metadata[i]['timepoint']: my_metadata[i][regression_variable.lower()] for i in range(0, len(my_metadata))}
egg_age_dict = {my_metadata[i]['timepoint']: my_metadata[i]['egg_age'] for i in range(0, len(my_metadata))}
# Exclude larval animals from ghost_age computation.
if regression_variable.lower() == 'ghost_age':
time_name = a_point.split(os.path.sep)[-1].split(' ')[0]
if egg_age_dict[time_name] > 0:
do_point = True
else:
do_point = False
if do_point:
my_time = a_point.split(os.path.sep)[-1].split(' ')[0]
my_age = metadata_dict[my_time]/24
mask_file = freeimage.read(a_point)
image_file = freeimage.read(a_point.replace('hmask.png', 'bf.png'))
my_samples = sample_8star(image_file, mask_file, 17, sample_number = samples_per_image)
my_codebook = np.zeros(texture_model.cluster_vectors.shape[0])
for a_sample in my_samples:
my_codebook[texture_model.classify_vector(a_sample)] += 1
my_codebook = my_codebook/np.sum(my_codebook)
histogram_array.append(my_codebook)
age_array.append(my_age)
return (histogram_array, age_array)
def colored_color_outlines(a_directory):
'''
For each outlined worm in a_directory, convert it to a grayscale 8-bit image with the outline in white and the rest of the image in black.
'''
for an_image in os.listdir(a_directory):
if 'new' not in an_image:
filepath = a_directory + os.path.sep + an_image
my_image = freeimage.read(filepath)
my_red = np.array([237, 28, 36])
is_red = np.abs(my_image[:, :, :3] - my_red)
is_red = (is_red.mean(axis = 2) < 1)
new_image = np.zeros(my_image.shape[:2]).astype('uint8')
new_image[is_red] = [-1]
freeimage.write(new_image, filepath.replace('.png', '_new.png'))
return
def corrected_fluorescence(image_location):
'''
Read in a fluroescence image corrected for flatfield and hot pixels.
'''
# Figure out where everything is.
split_pathsep = image_location.split(os.path.sep)
experiment_directory = os.path.sep.join(split_pathsep[:-2])
calibration_directory = experiment_directory + os.path.sep + 'calibrations'
time_point = split_pathsep[-1].split(' ')[0]
hot_threshold = 500
pickle_file = experiment_directory + os.path.sep + 'super_vignette.pickle'
if os.path.isfile(pickle_file):
with open(pickle_file, 'rb') as my_file:
super_vignette = pickle.load(my_file)
else:
raise BaseException('Missing super vignette!')
# Read in image and apply vignetting.
image_path = image_location
raw_image = freeimage.read(image_path)
raw_image[np.invert(super_vignette)] = 0
# Correct for flatfield.
flatfield_path = calibration_directory + os.path.sep + time_point + ' ' + 'fl_flatfield.tiff'
calibration_image = freeimage.read(flatfield_path)
corrected_image = raw_image*calibration_image
# Correct for hot pixels.
median_image = scipy.ndimage.filters.median_filter(corrected_image, size = 3)
difference_image = np.abs(corrected_image.astype('float64') - median_image.astype('float64')).astype('uint16')
hot_pixels = difference_image > hot_threshold
median_image_hot_pixels = median_image[hot_pixels]
corrected_image[hot_pixels] = median_image_hot_pixels
# Return the actual image.
return corrected_image
def _compress(self, image_path):
image_path = pathlib.Path(image_path)
assert image_path.suffix == '.png'
image = freeimage.read(image_path)
temp = tempfile.NamedTemporaryFile(dir=image_path.parent,
prefix=image_path.stem + 'compressing_', suffix='.png', delete=False)
try:
freeimage.write(image, temp.name, flags=self.level)
os.replace(temp.name, image_path)
except:
if os.path.exists(temp.name):
os.unlink(temp.name)
raise
finally:
temp.close()
def make_well_mask(out_dir, image_file, ignore_previous=False):
"""Calculate and store well mask if necessary.
Parameters:
out_dir: directory where well_mask.png should exist or be created.
image_file: path to an image to create the mask from, if it doesn't exist.
ignore_previous: if True, re-make mask even if it alredy exists on disk.
"""
out_dir = pathlib.Path(out_dir)
well_mask_f = out_dir / 'well_mask.png'
if ignore_previous or not well_mask_f.exists():
image = freeimage.read(image_file)
if image.dtype == numpy.uint16:
image = (image >> 8).astype(numpy.uint8)
well_mask = extract_wells.get_well_mask(image)
freeimage.write((well_mask * 255).astype(numpy.uint8), str(well_mask_f))
def measure_incyte_images(image_dir, image_glob='*FITC*'):
image_dir = pathlib.Path(image_dir)
assert image_dir.exists()
image_files = list(image_dir.glob(image_glob))
well_names = []
for image_file in sorted(image_files):
row = image_file.name[0] # first letter is row, then ' - ', then two-digit col
col = image_file.name[4:6]
well = row + col
well_names.append(well)
data_rows = []
for image_file in image_files:
print(str(image_file))
image = freeimage.read(image_file)
worm_mask = find_worm.find_worm_from_fluorescence(image)
data_rows.append(measure_fluorescence(image, worm_mask)[0])
write_measures(data_rows, well_names, image_dir.with_suffix('.csv'))
def _computeFocusMeasures(bgs, im_fpath, measure_mask, compute_measures, write_models, write_deltas, write_masks):
try:
im = freeimage.read(str(im_fpath))
except:
return
if bgs.model is not None:
# NB: Model and delta are written as float32 tiffs
try:
if write_models:
freeimage.write(
bgs.model,
str(im_fpath.parent / "{} wz_bgs_model.tiff".format(im_fpath.stem)),
freeimage.IO_FLAGS.TIFF_DEFLATE,
)
delta = numpy.abs(bgs.queryModelDelta(im))
if write_deltas:
freeimage.write(
delta,
str(im_fpath.parent / "{} wz_bgs_model_delta.tiff".format(im_fpath.stem)),
freeimage.IO_FLAGS.TIFF_DEFLATE,
)
mask = bgs.queryModelMask(im, delta)
antimask = mask == 0
if write_masks:
freeimage.write(
(mask * 255).astype(numpy.uint8),
str(im_fpath.parent / "{} wz_bgs_model_mask.png".format(im_fpath.stem)),
)
except:
return
if compute_measures:
focus_measures = {}
focus_measures["whole_image_hp_brenner_sum_of_squares"], focus_measures[
"whole_image_bp_brenner_sum_of_squares"
] = MaskedMultiBrenner((2560, 2160)).metric(im, measure_mask)
model_delta_squares = delta.astype(numpy.float64) ** 2
model_delta_squares[~measure_mask] = 0
focus_measures["model_delta_sum_of_squares"] = model_delta_squares.sum()
focus_measures["model_mask_count"] = mask.sum()
model_delta_squares[antimask] = 0
focus_measures["model_mask_region_delta_sum_of_squares"] = model_delta_squares.sum()
focus_measures["model_mask_region_image_hp_brenner_sum_of_squares"], focus_measures[
"model_mask_region_image_bp_brenner_sum_of_squares"
] = MaskedMultiBrenner((2560, 2160)).metric(im, mask)
return focus_measures