def test_gn_fit_fix_mu_sigma():
""" Test the fitting routine with fixed mean and std. """
# Create a T2WModality object
currdir = os.path.dirname(os.path.abspath(__file__))
path_data_t2w = os.path.join(currdir, 'data', 't2w')
t2w_mod = T2WModality(path_data_t2w)
t2w_mod.read_data_from_path()
# Create the GTModality object
path_data_gt = os.path.join(currdir, 'data', 'gt_folders')
path_data_gt_list = [os.path.join(path_data_gt, 'prostate'),
os.path.join(path_data_gt, 'pz'),
os.path.join(path_data_gt, 'cg'),
os.path.join(path_data_gt, 'cap')]
label_gt = ['prostate', 'pz', 'cg', 'cap']
gt_mod = GTModality()
gt_mod.read_data_from_path(cat_gt=label_gt, path_data=path_data_gt_list)
params = {'mu': 200., 'sigma': 70.}
gaussian_norm = GaussianNormalization(T2WModality(), params=params)
gaussian_norm.fit(t2w_mod, gt_mod, label_gt[0])
assert_almost_equal(gaussian_norm.fit_params_['mu'], 245.90,
decimal=DECIMAL_PRECISON)
assert_almost_equal(gaussian_norm.fit_params_['sigma'], 74.31,
decimal=DECIMAL_PRECISON)
def test_gn_fit_wt_gt():
""" Test the fitting routine without ground-truth. """
# Create a T2WModality object
currdir = os.path.dirname(os.path.abspath(__file__))
path_data_t2w = os.path.join(currdir, 'data', 't2w')
t2w_mod = T2WModality(path_data_t2w)
t2w_mod.read_data_from_path()
gaussian_norm = GaussianNormalization(T2WModality())
gaussian_norm.fit(t2w_mod)
assert_almost_equal(gaussian_norm.fit_params_['mu'], -8.13,
decimal=DECIMAL_PRECISON)
assert_almost_equal(gaussian_norm.fit_params_['sigma'], 11.35,
decimal=DECIMAL_PRECISON)
def find_normalization_params(pat_t2w, pat_gt, label):
# Create the normalization object and load the model
t2w_norm = GaussianNormalization(T2WModality())
# Read the T2W
t2w_mod = T2WModality()
t2w_mod.read_data_from_path(pat_t2w)
# Read the GT
gt_mod = GTModality()
gt_mod.read_data_from_path(label, pat_gt)
# Find the normalization parameters
t2w_norm.fit(t2w_mod, ground_truth=gt_mod, cat=label[0])
return t2w_norm
def test_gn_normalize():
""" Test the normalize function. """
# Create a T2WModality object
currdir = os.path.dirname(os.path.abspath(__file__))
path_data_t2w = os.path.join(currdir, 'data', 't2w')
t2w_mod = T2WModality(path_data_t2w)
t2w_mod.read_data_from_path()
# Create the GTModality object
path_data_gt = os.path.join(currdir, 'data', 'gt_folders')
path_data_gt_list = [os.path.join(path_data_gt, 'prostate'),
os.path.join(path_data_gt, 'pz'),
os.path.join(path_data_gt, 'cg'),
os.path.join(path_data_gt, 'cap')]
label_gt = ['prostate', 'pz', 'cg', 'cap']
gt_mod = GTModality()
gt_mod.read_data_from_path(cat_gt=label_gt, path_data=path_data_gt_list)
# Store the data before the normalization
pdf_copy = t2w_mod.pdf_.copy()
data_copy = t2w_mod.data_.copy()
# Normalize the data
gaussian_norm = GaussianNormalization(T2WModality())
gaussian_norm.fit(t2w_mod, gt_mod, 'prostate')
t2w_mod = gaussian_norm.normalize(t2w_mod)
# Check that the data are equal to what they should be
assert_array_almost_equal(t2w_mod.data_, (data_copy - 245.90) / 74.31,
decimal=DECIMAL_PRECISON)
# Denormalize the data
t2w_mod = gaussian_norm.denormalize(t2w_mod)
# Check that the data are equal to the original data
data = np.load(os.path.join(currdir, 'data', 'data_denormalize.npy'))
assert_array_equal(t2w_mod.data_, data)
data = np.load(os.path.join(currdir, 'data', 'pdf_denormalize.npy'))
assert_array_equal(t2w_mod.pdf_, data)
def test_gn_fit_wrong_params():
""" Test either if an error is raised when the params change just before
fitting. """
# Create a T2WModality object
currdir = os.path.dirname(os.path.abspath(__file__))
path_data_t2w = os.path.join(currdir, 'data', 't2w')
t2w_mod = T2WModality(path_data_t2w)
t2w_mod.read_data_from_path()
# Create the GTModality object
path_data_gt = os.path.join(currdir, 'data', 'gt_folders')
path_data_gt_list = [os.path.join(path_data_gt, 'prostate'),
os.path.join(path_data_gt, 'pz'),
os.path.join(path_data_gt, 'cg'),
os.path.join(path_data_gt, 'cap')]
label_gt = ['prostate', 'pz', 'cg', 'cap']
gt_mod = GTModality()
gt_mod.read_data_from_path(cat_gt=label_gt, path_data=path_data_gt_list)
gaussian_norm = GaussianNormalization(T2WModality())
gaussian_norm.params = 'None'
assert_raises(ValueError, gaussian_norm.fit, t2w_mod, gt_mod, label_gt[0])
def test_gn_save_load():
""" Test the save and load function. """
# Create a T2WModality object
currdir = os.path.dirname(os.path.abspath(__file__))
path_data_t2w = os.path.join(currdir, 'data', 't2w')
t2w_mod = T2WModality(path_data_t2w)
t2w_mod.read_data_from_path()
# Create the GTModality object
path_data_gt = os.path.join(currdir, 'data', 'gt_folders')
path_data_gt_list = [os.path.join(path_data_gt, 'prostate'),
os.path.join(path_data_gt, 'pz'),
os.path.join(path_data_gt, 'cg'),
os.path.join(path_data_gt, 'cap')]
label_gt = ['prostate', 'pz', 'cg', 'cap']
gt_mod = GTModality()
gt_mod.read_data_from_path(cat_gt=label_gt, path_data=path_data_gt_list)
# Normalize the data
gaussian_norm = GaussianNormalization(T2WModality())
gaussian_norm.fit(t2w_mod, gt_mod, 'prostate')
# Store the normalization object
filename = os.path.join(currdir, 'data', 'gn_obj.p')
gaussian_norm.save_to_pickles(filename)
# Load the object
gn_2 = GaussianNormalization.load_from_pickles(filename)
# Check that the different variables are the same
assert_equal(type(gn_2.base_modality_), type(gaussian_norm.base_modality_))
assert_equal(gn_2.fit_params_['mu'], gaussian_norm.fit_params_['mu'])
assert_equal(gn_2.fit_params_['sigma'], gaussian_norm.fit_params_['sigma'])
assert_equal(gn_2.is_fitted_, gaussian_norm.is_fitted_)
assert_array_equal(gn_2.roi_data_, gaussian_norm.roi_data_)
# Read the normalization information
pat_chg = id_patient_list[id_p].lower().replace(' ', '_') + '_norm.p'
filename = os.path.join(path_rician, pat_chg)
t2w_norm = RicianNormalization.load_from_pickles(filename)
# Normalize the data
t2w_mod = t2w_norm.normalize(t2w_mod)
else:
# Gaussian Normalization
# Read the normalization information
pat_chg = id_patient_list[id_p].lower().replace(' ', '_') + '_norm.p'
filename = os.path.join(path_gaussian, pat_chg)
t2w_norm = GaussianNormalization.load_from_pickles(filename)
# Normalize the data
t2w_mod = t2w_norm.normalize(t2w_mod)
# Rescale the data on 8 bits
t2w_mod.data_ = ((t2w_mod.data_ - EXTREM[0]) * (255. /
(EXTREM[1] - EXTREM[0])))
# Update the histogram
t2w_mod.update_histogram()
# Create the different parameters for the filter bank
frequencies = np.linspace(0.05, 0.25, num=4, endpoint=True)
alphas = np.linspace(0., np.pi, num=4, endpoint=True)
gammas = np.linspace(0., 2. * np.pi, num=8, endpoint=True)