def normalize_model_data(self):
# get the id of the slices by vertebral level
id_by_level = {}
for dic_slice in self.slices:
level_int = int(round(dic_slice.level))
if level_int not in id_by_level.keys():
id_by_level[level_int] = [dic_slice.id]
else:
id_by_level[level_int].append(dic_slice.id)
# get the average median values by level:
list_gm_by_level = []
list_wm_by_level = []
list_min_by_level = []
list_max_by_level = []
list_indexes = []
for level, list_id_slices in id_by_level.items():
if level != 0:
list_med_gm = []
list_med_wm = []
list_min = []
list_max = []
# get median GM and WM values for all slices of the same level:
for id_slice in list_id_slices:
slice = self.slices[id_slice]
for gm in slice.gm_seg_M:
med_gm = np.median(slice.im_M[gm==1])
list_med_gm.append(med_gm)
for wm in slice.wm_seg_M:
med_wm = np.median(slice.im_M[wm == 1])
list_med_wm.append(med_wm)
list_min.append(min(slice.im_M.flatten()[slice.im_M.flatten().nonzero()]))
list_max.append(max(slice.im_M.flatten()))
list_gm_by_level.append(np.mean(list_med_gm))
list_wm_by_level.append(np.mean(list_med_wm))
list_min_by_level.append(min(list_min))
list_max_by_level.append(max(list_max))
list_indexes.append(level)
# add level 0 for images with no level (or level not in model)
# average GM and WM for all slices, get min and max of all slices
list_gm_by_level.append(np.mean(list_gm_by_level))
list_wm_by_level.append(np.mean(list_wm_by_level))
list_min_by_level.append(min(list_min_by_level))
list_max_by_level.append(max(list_max_by_level))
list_indexes.append(0)
# save average median values in a Panda data frame
data_intensities = {'GM': pd.Series(list_gm_by_level, index=list_indexes), 'WM': pd.Series(list_wm_by_level, index=list_indexes), 'MIN': pd.Series(list_min_by_level, index=list_indexes), 'MAX': pd.Series(list_max_by_level, index=list_indexes)}
self.intensities = pd.DataFrame(data_intensities)
# Normalize slices using dic values
for dic_slice in self.slices:
level_int = int(round(dic_slice.level))
av_gm_slice, av_wm_slice = average_gm_wm([dic_slice], bin=True)
norm_im_M = normalize_slice(dic_slice.im_M, av_gm_slice, av_wm_slice, self.intensities['GM'][level_int], self.intensities['WM'][level_int], val_min=self.intensities['MIN'][level_int], val_max=self.intensities['MAX'][level_int])
dic_slice.set(im_m=norm_im_M)
def normalize_model_data(self):
# get the id of the slices by vertebral level
id_by_level = {}
for dic_slice in self.slices:
level_int = int(round(dic_slice.level))
if level_int not in id_by_level.keys():
id_by_level[level_int] = [dic_slice.id]
else:
id_by_level[level_int].append(dic_slice.id)
# get the average median values by level:
list_gm_by_level = []
list_wm_by_level = []
list_min_by_level = []
list_max_by_level = []
list_indexes = []
for level, list_id_slices in id_by_level.items():
if level != 0:
list_med_gm = []
list_med_wm = []
list_min = []
list_max = []
# get median GM and WM values for all slices of the same level:
for id_slice in list_id_slices:
slice = self.slices[id_slice]
for gm in slice.gm_seg_M:
med_gm = np.median(slice.im_M[gm == 1])
list_med_gm.append(med_gm)
for wm in slice.wm_seg_M:
med_wm = np.median(slice.im_M[wm == 1])
list_med_wm.append(med_wm)
list_min.append(min(slice.im_M.flatten()[slice.im_M.flatten().nonzero()]))
list_max.append(max(slice.im_M.flatten()))
list_gm_by_level.append(np.mean(list_med_gm))
list_wm_by_level.append(np.mean(list_med_wm))
list_min_by_level.append(min(list_min))
list_max_by_level.append(max(list_max))
list_indexes.append(level)
# add level 0 for images with no level (or level not in model)
# average GM and WM for all slices, get min and max of all slices
list_gm_by_level.append(np.mean(list_gm_by_level))
list_wm_by_level.append(np.mean(list_wm_by_level))
list_min_by_level.append(min(list_min_by_level))
list_max_by_level.append(max(list_max_by_level))
list_indexes.append(0)
# save average median values in a Panda data frame
data_intensities = {'GM': pd.Series(list_gm_by_level, index=list_indexes), 'WM': pd.Series(list_wm_by_level, index=list_indexes), 'MIN': pd.Series(list_min_by_level, index=list_indexes), 'MAX': pd.Series(list_max_by_level, index=list_indexes)}
self.intensities = pd.DataFrame(data_intensities)
# Normalize slices using dic values
for dic_slice in self.slices:
level_int = int(round(dic_slice.level))
av_gm_slice, av_wm_slice = average_gm_wm([dic_slice], bin=True)
norm_im_M = normalize_slice(dic_slice.im_M, av_gm_slice, av_wm_slice, self.intensities['GM'][level_int], self.intensities['WM'][level_int], val_min=self.intensities['MIN'][level_int], val_max=self.intensities['MAX'][level_int])
dic_slice.set(im_m=norm_im_M)
def label_fusion(self, list_dic_indexes_by_slice):
for target_slice in self.target_im:
# get list of slices corresponding to the indexes
list_dic_slices = [self.model.slices[j] for j in list_dic_indexes_by_slice[target_slice.id]]
# average slices GM and WM
# WM is not used anymore here, but the average_gm_wm() function is used in other parts of the code that need both the GM and WM averages
data_mean_gm, data_mean_wm = average_gm_wm(list_dic_slices)
# set negative values to 0
data_mean_gm[data_mean_gm < 0] = 0
# store segmentation into target_im
target_slice.set(gm_seg_m=data_mean_gm)
def label_fusion(self, list_dic_indexes_by_slice):
for target_slice in self.target_im:
# get list of slices corresponding to the indexes
list_dic_slices = [self.model.slices[j] for j in list_dic_indexes_by_slice[target_slice.id]]
# average slices GM and WM
# WM is not used anymore here, but the average_gm_wm() function is used in other parts of the code that need both the GM and WM averages
data_mean_gm, data_mean_wm = average_gm_wm(list_dic_slices)
# set negative values to 0
data_mean_gm[data_mean_gm < 0] = 0
# store segmentation into target_im
target_slice.set(gm_seg_m=data_mean_gm)
def label_fusion(self, list_dic_indexes_by_slice):
for target_slice in self.target_im:
# get list of slices corresponding to the indexes
list_dic_slices = [self.model.slices[j] for j in list_dic_indexes_by_slice[target_slice.id]]
# average slices GM and WM
data_mean_gm, data_mean_wm = average_gm_wm(list_dic_slices)
if self.param_seg.type_seg == 'bin':
# binarize GM seg
data_mean_gm[data_mean_gm >= 0.5] = 1
data_mean_gm[data_mean_gm < 0.5] = 0
# binarize WM seg
data_mean_wm[data_mean_wm >= 0.5] = 1
data_mean_wm[data_mean_wm < 0.5] = 0
# store segmentation into target_im
target_slice.set(gm_seg_m=data_mean_gm, wm_seg_m=data_mean_wm)
def get_gm_wm_by_level(self):
gm_seg_model = {} # dic of mean gm seg by vertebral level
wm_seg_model = {} # dic of mean wm seg by vertebral level
# get id of the slices by level
slices_by_level = {}
for dic_slice in self.slices:
level_int = int(round(dic_slice.level))
if level_int not in slices_by_level.keys():
slices_by_level[level_int] = [dic_slice]
else:
slices_by_level[level_int].append(dic_slice)
# get average gm and wm by level
for level, list_slices in slices_by_level.items():
data_mean_gm, data_mean_wm = average_gm_wm(list_slices)
gm_seg_model[level] = data_mean_gm
wm_seg_model[level] = data_mean_wm
# for level=0 (no leve or level not in model) output average GM and WM seg across all model data
gm_seg_model[0] = np.mean(gm_seg_model.values(), axis=0)
wm_seg_model[0] = np.mean(wm_seg_model.values(), axis=0)
return gm_seg_model, wm_seg_model
def get_gm_wm_by_level(self):
gm_seg_model = {} # dic of mean gm seg by vertebral level
wm_seg_model = {} # dic of mean wm seg by vertebral level
# get id of the slices by level
slices_by_level = {}
for dic_slice in self.slices:
level_int = int(np.round(dic_slice.level))
if level_int not in set(slices_by_level.keys()):
slices_by_level[level_int] = [dic_slice]
else:
slices_by_level[level_int].append(dic_slice)
# get average gm and wm by level
for level, list_slices in slices_by_level.items():
data_mean_gm, data_mean_wm = average_gm_wm(list_slices)
gm_seg_model[level] = data_mean_gm
wm_seg_model[level] = data_mean_wm
# for level=0 (no leve or level not in model) output average GM and WM seg across all model data
gm_seg_model[0] = np.mean(list(gm_seg_model.values()), axis=0)
wm_seg_model[0] = np.mean(list(wm_seg_model.values()), axis=0)
return gm_seg_model, wm_seg_model
