Skip to content

Commit

Permalink
median ob implemented with unit tests. this refs #9
Browse files Browse the repository at this point in the history
  • Loading branch information
@JeanBilheux
JeanBilheux committed Mar 21, 2022
1 parent fc44066 commit f1f2b00
Show file tree
Hide file tree
Showing 2 changed files with 170 additions and 102 deletions.
57 changes: 44 additions & 13 deletions NeuNorm/normalization.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -123,7 +123,7 @@ def load(self, file='', folder='', data=None, data_type='sample', auto_gamma_fil
takes_its_going_to_take = self.calculate_how_long_its_going_to_take(index_we_are=_index + 1,
time_it_took_so_far=end_time - start_time,
total_number_of_loop=len(
file))
file))
time_remaining_ui.value = "{}".format(takes_its_going_to_take)

if notebook:
Expand Down Expand Up @@ -160,7 +160,7 @@ def load(self, file='', folder='', data=None, data_type='sample', auto_gamma_fil
takes_its_going_to_take = self.calculate_how_long_its_going_to_take(index_we_are=_index + 1,
time_it_took_so_far=end_time - start_time,
total_number_of_loop=len(
list_images))
list_images))
time_remaining_ui.value = "{}".format(takes_its_going_to_take)

if notebook:
Expand Down Expand Up @@ -412,18 +412,21 @@ def save_or_check_shape(self, data=None, data_type='sample'):
if (not (_prev_width == width)) or (not (_prev_height == height)):
raise IOError("Shape of {} do not match previous loaded data set!".format(data_type))

def normalization(self, roi=None, force=False, force_mean_ob=False, notebook=False, use_only_sample=False):
def normalization(self, roi=None, force=False, force_mean_ob=False, force_median_ob=False, notebook=False,
use_only_sample=False):
"""normalization of the data
Parameters:
roi: ROI object or list of ROI objects - object defines the region of the sample and OB that have to match
in intensity
in intensity
force: boolean - True will force the normalization to occur, even if it had been
run before with the same data set (default False)
notebook: boolean - turn on this option if you run the library from a
notebook to have a progress bar displayed showing you the progress of the loading (default False)
use_only_sample - turn on this option to normalize the sample data using the ROI on the sample. each pixel
counts will be divided by the average counts of all the ROI of the same image
force_mean_ob: boolean - True will force using only 1 OB that is the mean of all the OBs
force_median_ob: boolean - True will force using only 1 OB that is the median of all the OBs
notebook: boolean - turn on this option if you run the library from a
notebook to have a progress bar displayed showing you the progress of the loading (default False)
use_only_sample: turn on this option to normalize the sample data using the ROI on the sample. each pixel
counts will be divided by the average counts of all the ROI of the same image
Return:
True - status of the normalization (True if every went ok, this is mostly used for the unit test)
Expand All @@ -435,6 +438,7 @@ def normalization(self, roi=None, force=False, force_mean_ob=False, notebook=Fal
IOError: if size of sample and OB do not match
"""

if not force:
# does nothing if normalization has already been run
if self.__exec_process_status['normalization']:
Expand Down Expand Up @@ -480,9 +484,9 @@ def normalization(self, roi=None, force=False, force_mean_ob=False, notebook=Fal
_x1 = roi.x1
_y1 = roi.y1

_sample_corrected_normalized = [_sample / np.mean(_sample[_y0:_y1 + 1, _x0:_x1 + 1])
_sample_corrected_normalized = [_sample / np.median(_sample[_y0:_y1 + 1, _x0:_x1 + 1])
for _sample in self.data['sample']['data']]
_ob_corrected_normalized = [_ob / np.mean(_ob[_y0:_y1 + 1, _x0:_x1 + 1])
_ob_corrected_normalized = [_ob / np.median(_ob[_y0:_y1 + 1, _x0:_x1 + 1])
for _ob in self.data['ob']['data']]

else:
Expand All @@ -493,10 +497,12 @@ def normalization(self, roi=None, force=False, force_mean_ob=False, notebook=Fal
self.data[DataType.sample]['data'] = _sample_corrected_normalized
self.data[DataType.ob]['data'] = _ob_corrected_normalized

# if the number of sample and ob do not match, use mean of obs
# if the number of sample and ob do not match, use median of obs
nbr_sample = len(self.data['sample']['file_name'])
nbr_ob = len(self.data['ob']['file_name'])
if (nbr_sample != nbr_ob) or force_mean_ob: # work with mean ob

# will be deprecated soon!
if force_mean_ob: # work with mean ob
_ob_corrected_normalized = np.mean(_ob_corrected_normalized, axis=0)
self.data['ob']['data_mean'] = _ob_corrected_normalized
_working_ob = copy.deepcopy(_ob_corrected_normalized)
Expand All @@ -521,6 +527,31 @@ def normalization(self, roi=None, force=False, force_mean_ob=False, notebook=Fal
if notebook:
w1.value = _index + 1

elif (nbr_sample != nbr_ob) or force_median_ob: # work with median ob
_ob_corrected_normalized = np.median(_ob_corrected_normalized, axis=0)
self.data['ob']['data_mean'] = _ob_corrected_normalized
_working_ob = copy.deepcopy(_ob_corrected_normalized)
_working_ob[_working_ob == 0] = np.NaN

if notebook:
# turn on progress bar
_message = "Normalization"
box1 = widgets.HBox([widgets.Label(_message,
layout=widgets.Layout(width='20%')),
widgets.IntProgress(max=len(self.data['sample']['data']))])
display(box1)
w1 = box1.children[1]

normalized_data = []
for _index, _sample in enumerate(self.data['sample']['data']):
_norm = np.divide(_sample, _working_ob)
_norm[np.isnan(_norm)] = 0
_norm[np.isinf(_norm)] = 0
normalized_data.append(_norm)

if notebook:
w1.value = _index + 1

else: # 1 ob for each sample
# produce normalized data
sample_ob = zip(self.data['sample']['data'], self.data['ob']['data'])
Expand Down Expand Up @@ -584,7 +615,7 @@ def calculate_corrected_normalized_without_ob(self, roi=None):
_x1 = roi.x1
_y1 = roi.y1

normalized_data = [_sample / np.mean(_sample[_y0:_y1 + 1, _x0:_x1 + 1])
normalized_data = [_sample / np.median(_sample[_y0:_y1 + 1, _x0:_x1 + 1])
for _sample in self.data['sample']['data']]

return normalized_data
Expand Down
Loading

0 comments on commit f1f2b00

Please sign in to comment.