def make_workspace(self, image, mask=None, labels=None, binary_image=None):
'''Make a workspace and IdentifyPrimaryObjects module
image - the intensity image for thresholding
mask - if present, the "don't analyze" mask of the intensity image
labels - if thresholding per-object, the labels matrix needed
binary_image - if thresholding using a binary image, the image
'''
module = YS.IdentifyYeastCells()
module.module_num = 1
module.input_image_name.value = IMAGE_NAME
module.object_name.value = OBJECTS_NAME
module.binary_image.value = BINARY_IMAGE_NAME
module.masking_objects.value = MASKING_OBJECTS_NAME
pipeline = cellprofiler.pipeline.Pipeline()
pipeline.add_module(module)
m = cpmeas.Measurements()
cpimage = cpi.Image(image, mask=mask)
m.add(IMAGE_NAME, cpimage)
if binary_image is not None:
m.add(BINARY_IMAGE_NAME, cpi.Image(binary_image))
object_set = cpo.ObjectSet()
if labels is not None:
o = cpo.Objects()
o.segmented = labels
object_set.add_objects(o, MASKING_OBJECTS_NAME)
workspace = cellprofiler.workspace.Workspace(pipeline, module, m,
object_set, m, None)
return workspace, module
def test_02_02_discard_small(self):
x = YS.IdentifyYeastCells()
x.object_name.value = OBJECTS_NAME
x.segmentation_precision.value = 11
x.input_image_name.value = IMAGE_NAME
x.background_brighter_then_cell_inside.value = False
x.average_cell_diameter.value = 30
x.min_cell_area.value = 500
x.max_cell_area.value = 5000
x.advanced_cell_filtering.value = True
img = np.ones((200, 200)) * 0.5
draw_brightfield_cell(img, 100, 100, 20, False)
draw_brightfield_cell(img, 25, 25, 10, False)
draw_brightfield_cell(img, 150, 150, 15, False)
image = cpi.Image(img, file_name="test_02_02_discard_small")
image_set_list = cpi.ImageSetList()
image_set = image_set_list.get_image_set(0)
image_set.providers.append(cpi.VanillaImageProvider(IMAGE_NAME, image))
object_set = cpo.ObjectSet()
measurements = cpmeas.Measurements()
pipeline = cellprofiler.pipeline.Pipeline()
x.run(Workspace(pipeline, x, image_set, object_set, measurements,
None))
objects = object_set.get_objects(OBJECTS_NAME)
self.assertEqual(objects.segmented[25, 25] > 0, 0,
"The small object was not filtered out")
self.assertEqual(objects.segmented[150, 150] > 0, 1,
"The medium object was not there")
self.assertEqual(objects.segmented[100, 100] > 0, 1,
"The large object was not there")
location_center_x = measurements.get_current_measurement(
OBJECTS_NAME, "Location_Center_X")
self.assertTrue(isinstance(location_center_x, np.ndarray))
self.assertEqual(np.product(location_center_x.shape), 2)
def test_01_07_extreme_params(self):
x = YS.IdentifyYeastCells()
x.object_name.value = OBJECTS_NAME
x.segmentation_precision.value = 14
x.input_image_name.value = IMAGE_NAME
x.background_brighter_then_cell_inside.value = False
x.average_cell_diameter.value = 77
img = get_two_cell_mask()
draw_disc(img, (5, 5), 2, 0.7)
draw_disc(img, (35, 11), 3, 0.2)
img = convert_to_brightfield(img, False)
image = cpi.Image(img, file_name="test_01_07_extreme_params")
image_set_list = cpi.ImageSetList()
image_set = image_set_list.get_image_set(0)
image_set.providers.append(cpi.VanillaImageProvider(IMAGE_NAME, image))
object_set = cpo.ObjectSet()
measurements = cpmeas.Measurements()
pipeline = cellprofiler.pipeline.Pipeline()
x.run(Workspace(pipeline, x, image_set, object_set, measurements,
None))
objects = object_set.get_objects(OBJECTS_NAME)
segmented = objects.segmented
self.assertTrue(np.all(
segmented == 0)) # no found because of parameters (no foreground)
self.assertEqual(0, objects.count)
def test_01_06_fill_holes(self):
x = YS.IdentifyYeastCells()
x.object_name.value = OBJECTS_NAME
x.segmentation_precision.value = 11
x.input_image_name.value = IMAGE_NAME
x.background_brighter_then_cell_inside.value = False
x.average_cell_diameter.value = 5
img = np.zeros((40, 40))
draw_disc(img, (10, 10), 7, .5)
draw_disc(img, (30, 30), 7, .5)
img[10, 10] = 0
img[30, 30] = 0
image = cpi.Image(img, file_name="test_01_06_fill_holes")
image_set_list = cpi.ImageSetList()
image_set = image_set_list.get_image_set(0)
image_set.providers.append(cpi.VanillaImageProvider(IMAGE_NAME, image))
object_set = cpo.ObjectSet()
measurements = cpmeas.Measurements()
pipeline = cellprofiler.pipeline.Pipeline()
x.run(Workspace(pipeline, x, image_set, object_set, measurements,
None))
objects = object_set.get_objects(OBJECTS_NAME)
self.assertTrue(objects.segmented[10, 10] > 0)
self.assertTrue(objects.segmented[30, 30] > 0)
def test_01_01_test_one_object(self):
x = YS.IdentifyYeastCells()
x.object_name.value = OBJECTS_NAME
x.segmentation_precision.value = 11
x.input_image_name.value = IMAGE_NAME
x.background_brighter_then_cell_inside.value = False
x.average_cell_diameter.value = 10
img = convert_to_brightfield(get_one_cell_mask(), False)
image = cpi.Image(img, file_name="test_01_01_test_one_object")
image_set_list = cpi.ImageSetList()
image_set = image_set_list.get_image_set(0)
image_set.providers.append(cpi.VanillaImageProvider(IMAGE_NAME, image))
object_set = cpo.ObjectSet()
measurements = cpmeas.Measurements()
pipeline = cellprofiler.pipeline.Pipeline()
x.run(Workspace(pipeline, x, image_set, object_set, measurements,
None))
self.assertEqual(len(object_set.object_names), 1)
self.assertTrue(OBJECTS_NAME in object_set.object_names)
objects = object_set.get_objects(OBJECTS_NAME)
segmented = objects.segmented
self.assertTrue(is_segmentation_correct(get_one_cell_mask(),
segmented))
self.assertTrue("Image" in measurements.get_object_names())
self.assertTrue(OBJECTS_NAME in measurements.get_object_names())
self.assertTrue(
"Features_Quality" in measurements.get_feature_names(OBJECTS_NAME))
quality = measurements.get_current_measurement(OBJECTS_NAME,
"Features_Quality")
self.assertTrue(quality[0] > 0)
self.assertTrue("Location_Center_Y" in measurements.get_feature_names(
OBJECTS_NAME))
location_center_y = measurements.get_current_measurement(
OBJECTS_NAME, "Location_Center_Y")
self.assertTrue(isinstance(location_center_y, np.ndarray))
self.assertEqual(np.product(location_center_y.shape), 1)
self.assertTrue(location_center_y[0] > 8)
self.assertTrue(location_center_y[0] < 12)
self.assertTrue("Location_Center_X" in measurements.get_feature_names(
OBJECTS_NAME))
location_center_x = measurements.get_current_measurement(
OBJECTS_NAME, "Location_Center_X")
self.assertTrue(isinstance(location_center_x, np.ndarray))
self.assertEqual(np.product(location_center_x.shape), 1)
self.assertTrue(location_center_x[0] > 13)
self.assertTrue(location_center_x[0] < 16)
columns = x.get_measurement_columns(pipeline)
for object_name in (cpmeas.IMAGE, OBJECTS_NAME):
ocolumns = [x for x in columns if x[0] == object_name]
features = measurements.get_feature_names(object_name)
self.assertEqual(len(ocolumns), len(features))
self.assertTrue(all([column[1] in features
for column in ocolumns]))
def test_02_04_mask_input_image(self):
x = YS.IdentifyYeastCells()
x.object_name.value = OBJECTS_NAME
x.input_image_name.value = IMAGE_NAME
x.segmentation_precision.value = 11
x.background_brighter_then_cell_inside.value = False
x.average_cell_diameter.value = 30
img = np.ones((200, 200)) * 0.5
draw_brightfield_cell(img, 100, 100, 20, False)
draw_brightfield_cell(img, 25, 25, 10, False)
img[0:10, 0:10] = 1
img[180:200, 180:200] = 0
msk = np.zeros((200, 200))
msk[0:10, 0:10] = 1
msk[180:200, 180:200] = 1
image = cpi.Image(img, file_name="test_02_04_mask_input_image")
mask = cpi.Image(msk)
image_set_list = cpi.ImageSetList()
image_set = image_set_list.get_image_set(0)
image_set.providers.append(cpi.VanillaImageProvider(IMAGE_NAME, image))
image_set.providers.append(
cpi.VanillaImageProvider(MASK_IMAGE_NAME, mask))
# first try without masking
object_set = cpo.ObjectSet()
measurements = cpmeas.Measurements()
pipeline = cellprofiler.pipeline.Pipeline()
x.run(Workspace(pipeline, x, image_set, object_set, measurements,
None))
objects = object_set.get_objects(OBJECTS_NAME)
self.assertEqual(0, objects.segmented.max(),
"Cells should not be found due to the distractors")
# now if we use masking option we should find these cells
object_set = cpo.ObjectSet()
measurements = cpmeas.Measurements()
pipeline = cellprofiler.pipeline.Pipeline()
x.ignore_mask_image_name.value = MASK_IMAGE_NAME
x.run(Workspace(pipeline, x, image_set, object_set, measurements,
None))
objects = object_set.get_objects(OBJECTS_NAME)
self.assertEqual(objects.segmented[25, 25] > 0, 1,
"The small object was not there")
self.assertEqual(objects.segmented[100, 100] > 0, 1,
"The large object was not there")
def test_01_05_test_two_flu_dark_objects(self):
x = YS.IdentifyYeastCells()
x.object_name.value = OBJECTS_NAME
x.segmentation_precision.value = 11
x.input_image_name.value = IMAGE_NAME
x.background_brighter_then_cell_inside.value = True
x.bright_field_image.value = False
x.average_cell_diameter.value = 10
img = convert_to_fluorescent(get_two_cell_mask(), True)
image = cpi.Image(img,
file_name="test_01_05_test_two_flu_dark_objects")
image_set_list = cpi.ImageSetList()
image_set = image_set_list.get_image_set(0)
image_set.providers.append(cpi.VanillaImageProvider(IMAGE_NAME, image))
object_set = cpo.ObjectSet()
measurements = cpmeas.Measurements()
pipeline = cellprofiler.pipeline.Pipeline()
x.run(Workspace(pipeline, x, image_set, object_set, measurements,
None))
self.assertEqual(len(object_set.object_names), 1)
self.assertTrue(OBJECTS_NAME in object_set.object_names)
objects = object_set.get_objects(OBJECTS_NAME)
self.assertTrue(
is_segmentation_correct(get_two_cell_mask(), objects.segmented))
self.assertTrue("Image" in measurements.get_object_names())
self.assertTrue(OBJECTS_NAME in measurements.get_object_names())
self.assertTrue(
"Features_Quality" in measurements.get_feature_names(OBJECTS_NAME))
quality = measurements.get_current_measurement(OBJECTS_NAME,
"Features_Quality")
self.assertTrue(len(quality) == 2)
self.assertTrue("Location_Center_X" in measurements.get_feature_names(
OBJECTS_NAME))
location_center_y = measurements.get_current_measurement(
OBJECTS_NAME, "Location_Center_Y")
location_center_x = measurements.get_current_measurement(
OBJECTS_NAME, "Location_Center_X")
positions = sorted(zip(location_center_x, location_center_y))
self.assertEqual(2, len(positions))
self.assertRange(3, 18, positions[0][0])
self.assertRange(25, 45, positions[0][1])
self.assertRange(20, 40, positions[1][0])
self.assertRange(5, 25, positions[1][1])
def test_01_00_test_zero_objects(self):
x = YS.IdentifyYeastCells()
x.object_name.value = OBJECTS_NAME
x.segmentation_precision.value = 11
x.input_image_name.value = IMAGE_NAME
x.background_brighter_then_cell_inside.value = False
x.average_cell_diameter.value = 5
img = np.zeros((25, 25))
image = cpi.Image(img, file_name="test_01_00_test_zero_objects")
image_set_list = cpi.ImageSetList()
image_set = image_set_list.get_image_set(0)
image_set.providers.append(cpi.VanillaImageProvider(IMAGE_NAME, image))
object_set = cpo.ObjectSet()
measurements = cpmeas.Measurements()
pipeline = cellprofiler.pipeline.Pipeline()
x.run(Workspace(pipeline, x, image_set, object_set, measurements,
None))
self.assertEqual(len(object_set.object_names), 1)
self.assertTrue(OBJECTS_NAME in object_set.object_names)
objects = object_set.get_objects(OBJECTS_NAME)
segmented = objects.segmented
self.assertTrue(np.all(segmented == 0))
self.assertTrue("Image" in measurements.get_object_names())
self.assertTrue(OBJECTS_NAME in measurements.get_object_names())
self.assertTrue(
"Count_" + OBJECTS_NAME in measurements.get_feature_names("Image"))
count = measurements.get_current_measurement("Image",
"Count_" + OBJECTS_NAME)
self.assertEqual(count, 0)
self.assertTrue("Location_Center_X" in measurements.get_feature_names(
OBJECTS_NAME))
location_center_x = measurements.get_current_measurement(
OBJECTS_NAME, "Location_Center_X")
self.assertTrue(isinstance(location_center_x, np.ndarray))
self.assertEqual(np.product(location_center_x.shape), 0)
self.assertTrue("Location_Center_Y" in measurements.get_feature_names(
OBJECTS_NAME))
location_center_y = measurements.get_current_measurement(
OBJECTS_NAME, "Location_Center_Y")
self.assertTrue(isinstance(location_center_y, np.ndarray))
self.assertEqual(np.product(location_center_y.shape), 0)
def test_02_03_use_background_image(self):
x = YS.IdentifyYeastCells()
x.object_name.value = OBJECTS_NAME
x.input_image_name.value = IMAGE_NAME
x.segmentation_precision.value = 11
x.background_image_name.value = BACKGROUND_IMAGE_NAME
x.background_elimination_strategy.value = YS.BKG_FILE
x.background_brighter_then_cell_inside.value = False
x.average_cell_diameter.value = 30
img = np.ones((200, 200)) * 0.5
draw_brightfield_cell(img, 100, 100, 20, False)
draw_brightfield_cell(img, 25, 25, 10, False)
draw_brightfield_cell(img, 150, 150, 15, False) # background blob
bkg = np.ones((200, 200)) * 0.5
draw_brightfield_cell(bkg, 150, 150, 15, False) # background blob
image = cpi.Image(img, file_name="test_02_03_use_background_image")
background = cpi.Image(bkg)
image_set_list = cpi.ImageSetList()
image_set = image_set_list.get_image_set(0)
image_set.providers.append(cpi.VanillaImageProvider(IMAGE_NAME, image))
image_set.providers.append(
cpi.VanillaImageProvider(BACKGROUND_IMAGE_NAME, background))
object_set = cpo.ObjectSet()
measurements = cpmeas.Measurements()
pipeline = cellprofiler.pipeline.Pipeline()
x.run(Workspace(pipeline, x, image_set, object_set, measurements,
None))
objects = object_set.get_objects(OBJECTS_NAME)
self.assertEqual(objects.segmented[25, 25] > 0, 1,
"The small object was not there")
self.assertEqual(objects.segmented[100, 100] > 0, 1,
"The large object was not there")
self.assertEqual(objects.segmented[150, 150] > 0, 0,
"The background blob was not filtered out")
def test_00_00_init(self):
x = YS.IdentifyYeastCells()
def test_03_02_fitting_background_masked(self):
# reduce depth of fitting to speed up testing
import cellstar.parameter_fitting.pf_process as process
import cellstar.parameter_fitting.pf_runner as runner
process.SEARCH_LENGTH_NORMAL = 20
# test one core
process.get_max_workers = lambda: 1
runner.get_max_workers = lambda: 1
# make it deterministic
np.random.seed(1)
x = YS.IdentifyYeastCells()
x.object_name.value = OBJECTS_NAME
x.input_image_name.value = IMAGE_NAME
x.segmentation_precision.value = 9 # so that it is faster for fitting
x.maximal_cell_overlap.value = 0.4
x.background_brighter_then_cell_inside.value = False
x.average_cell_diameter.value = 12
x.autoadaptation_steps.value = 1
x.background_image_name.value = BACKGROUND_IMAGE_NAME
x.background_elimination_strategy.value = YS.BKG_FILE
x.ignore_mask_image_name.value = MASK_IMAGE_NAME
img = np.ones((50, 50)) * 0.5
draw_brightfield_cell(img, 7, 7, 5, False)
draw_brightfield_cell(img, 25, 28, 5, False)
draw_brightfield_cell(img, 15, 16, 5, False)
draw_brightfield_cell(img, 40, 40, 4, False)
draw_disc(img, (7, 7), 5, .65)
draw_disc(img, (25, 28), 5, .65)
draw_disc(img, (15, 16), 5, .65)
draw_disc(img, (40, 40), 4, .65)
img = img + np.random.normal(0.5, 0.01, img.shape)
img = scipy.ndimage.gaussian_filter(img, 2)
# bright flares
draw_disc(img, (40, 10), 10, 1.5)
ignore_mask = np.zeros((50, 50), dtype=bool)
draw_disc(ignore_mask, (40, 10), 10, 1)
# dark areas
draw_disc(img, (50, 25), 10, 0.0)
background_mask = np.ones((50, 50)) * 0.5
draw_disc(background_mask, (50, 25), 10, 0.0)
label = np.zeros((50, 50), dtype=int)
draw_disc(label, (10, 10), 7, 1)
draw_disc(label, (25, 20), 7, 2)
image = cpi.Image(img,
file_name="test_03_02_fitting_background_masked")
mask = cpi.Image(ignore_mask)
background = cpi.Image(background_mask)
image_set_list = cpi.ImageSetList()
image_set = image_set_list.get_image_set(0)
image_set.providers.append(cpi.VanillaImageProvider(IMAGE_NAME, image))
image_set.providers.append(
cpi.VanillaImageProvider(MASK_IMAGE_NAME, mask))
image_set.providers.append(
cpi.VanillaImageProvider(BACKGROUND_IMAGE_NAME, background))
old_params = ast.literal_eval(x.autoadapted_params.value)
input_processed, background_processed, ignore_mask_processed = x.preprocess_images(
img, background_mask, ignore_mask)
x.fit_parameters(input_processed, background_processed,
ignore_mask_processed, label,
x.autoadaptation_steps.value * 2, lambda x: True,
lambda secs: time.sleep(secs))
new_params = ast.literal_eval(x.autoadapted_params.value)
self.assertNotEqual(old_params[0], new_params[0])
self.assertNotEqual(old_params[1], new_params[1])
# now if we use new parameters option we should find these cells
object_set = cpo.ObjectSet()
measurements = cpmeas.Measurements()
pipeline = cellprofiler.pipeline.Pipeline()
x.segmentation_precision.value = 11
x.run(Workspace(pipeline, x, image_set, object_set, measurements,
None))
objects = object_set.get_objects(OBJECTS_NAME)
# 3 or four objects are acceptable
self.assertLessEqual(3, objects.segmented.max())
self.assertLessEqual(objects.segmented.max(), 4)
colours = sorted([
objects.segmented[10, 10], objects.segmented[25, 25],
objects.segmented[40, 40]
])
self.assertEqual(colours[0], 1)
self.assertEqual(colours[1], 2)
self.assertEqual(colours[2], 3)
def test_03_01_simple_fitting(self):
# reduce depth of fitting to speed up testing
import cellstar.parameter_fitting.pf_process as process
import cellstar.parameter_fitting.pf_runner as runner
process.SEARCH_LENGTH_NORMAL = 20
# test multicore but only two cores
process.get_max_workers = lambda: 2
runner.get_max_workers = lambda: 2
# make it deterministic
np.random.seed(1)
x = YS.IdentifyYeastCells()
x.object_name.value = OBJECTS_NAME
x.input_image_name.value = IMAGE_NAME
x.segmentation_precision.value = 9 # so that it is faster for fitting
x.maximal_cell_overlap.value = 0.4
x.background_brighter_then_cell_inside.value = False
x.average_cell_diameter.value = 30
x.autoadaptation_steps.value = 1
img = np.ones((200, 200)) * 0.5
draw_brightfield_cell(img, 100, 100, 15, False)
draw_brightfield_cell(img, 120, 120, 15, False)
draw_brightfield_cell(img, 110, 70, 15, False)
draw_brightfield_cell(img, 160, 160, 10, False)
draw_disc(img, (100, 100), 15, .65)
draw_disc(img, (120, 120), 15, .65)
draw_disc(img, (110, 70), 15, .65)
draw_disc(img, (160, 160), 10, .65)
img = img + np.random.normal(3., 0.01, img.shape)
img = scipy.ndimage.gaussian_filter(img, 3)
label = np.zeros((200, 200), dtype=int)
draw_disc(label, (100, 100), 15, 1)
draw_disc(label, (110, 70), 15, 2)
image = cpi.Image(img, file_name="test_03_01_simple_fitting")
image_set_list = cpi.ImageSetList()
image_set = image_set_list.get_image_set(0)
image_set.providers.append(cpi.VanillaImageProvider(IMAGE_NAME, image))
old_params = ast.literal_eval(x.autoadapted_params.value)
input_processed, background_processed, ignore_mask_processed = x.preprocess_images(
img, None, None)
x.fit_parameters(input_processed, background_processed,
ignore_mask_processed, label,
x.autoadaptation_steps.value * 2, lambda x: True,
lambda secs: time.sleep(secs))
new_params = ast.literal_eval(x.autoadapted_params.value)
self.assertNotEqual(old_params[0], new_params[0])
self.assertNotEqual(old_params[1], new_params[1])
# now if we use new parameters option we should find these cells
object_set = cpo.ObjectSet()
measurements = cpmeas.Measurements()
pipeline = cellprofiler.pipeline.Pipeline()
x.segmentation_precision.value = 11
x.run(Workspace(pipeline, x, image_set, object_set, measurements,
None))
objects = object_set.get_objects(OBJECTS_NAME)
self.assertEqual(4, objects.segmented.max())
colours = sorted([
objects.segmented[100, 100], objects.segmented[120, 120],
objects.segmented[110, 70], objects.segmented[160, 160]
])
self.assertEqual(colours[0], 1)
self.assertEqual(colours[1], 2)
self.assertEqual(colours[2], 3)
self.assertEqual(colours[3], 4)