def testMethodFatorizationPriorBound(self):
tes = np.array([0.07, 0.13])
tes = tes[:, np.newaxis]
t2 = np.array([0.1, 2])
f = np.array([0.7, 0.3])
(A, _) = utils.build_A(t2, tes, f)
b = 1000*np.ones(31)
b[0] = 0
Scsf = np.exp(np.dot(-b, 3e-3))
S = np.random.random(31)
S[0] = 1
S = np.stack((S, Scsf))
X = np.dot(A,S)
mask = 1
params = {'max_iters': 10, 'max_sources': 2, 'tolfun': 1e-8, 'tolx': 1e-8 }
t2_bounds = {'1': [0, 0.2], '2': [1.5, 2.5]}
s_prior = {'2': Scsf}
r = c = s = 1
result = self.__create_result_object()
obss = bss_cals.BssCals()
obss._method_factorization(X, tes, mask, params, result, r, c, s, s_prior, t2_bounds)
np.testing.assert_array_almost_equal(result.T2[r,c,s,:], t2, 6, 'Unexpected T2 values')
np.testing.assert_array_almost_equal(result.f[r,c,s,:], f, 6, 'Unexpected f values')
self.assertEqual(result.pd[r,c,s], 1, 'Unexpected S0 values')
np.testing.assert_array_almost_equal(result.sources[r,c,s,:,:], S, 3, 'Unexpected sources')
self.assertAlmostEqual(result.rel_error[r,c,s], 0, 6, 'Unexpected S0 values')
def testFixSNoPrior(self):
nsamples = 30
Sref = np.random.random((2, nsamples))
s_priors = None
obss = bss_cals.BssCals()
S = obss._fix_S(Sref, s_priors)
np.testing.assert_array_almost_equal(S, Sref, 6,'Wrong H')
def testBoundANoBounds(self):
t2_bounds = None
tes = np.array([0.07, 0.13])
tes = tes[:, None]
Aref = np.ones((2, 2))
A = np.ones((2, 2))
obss = bss_cals.BssCals()
A = obss._bound_A(A, t2_bounds, tes)
self.assertIsNotNone(A, 'Missing bounded A')
np.testing.assert_array_almost_equal(A, Aref, 6, 'Wrong A')
def testNegativeTolfun(self):
obss = bss_cals.BssCals()
dwi = ['data_TE1.nii.gz', 'data_TE2.nii.gz']
TE = [60, 120]
data = {'dwi': dwi, 'te': TE}
mask = 'mask.nii.gz'
params = {'max_iters': 100, 'max_sources': 2, 'tolfun': -1, 'tolx': 1e-6}
with self.assertRaises(bss.BssException) as context:
results = obss.factorize(data, mask, params,'/')
self.assertIsNone(results, 'Unexpected result')
self.assertTrue('Parameter tolfun must be >= 0.' in context.exception)
def testNoMaxIters(self):
obss = bss_cals.BssCals()
dwi = ['data_TE1.nii.gz', 'data_TE2.nii.gz']
TE = [60, 120]
data = {'dwi': dwi, 'te': TE}
mask = 'mask.nii.gz'
params = {'max_sources': 2}
with self.assertRaises(bss.BssException) as context:
results = obss.factorize(data, mask, params,'/')
self.assertIsNone(results, 'Unexpected result')
self.assertTrue('Missing max_iters parameter.' in context.exception)
def testFixSPrior(self):
nsamples = 30
S = np.random.random((2, nsamples))
s_prior = np.random.random(nsamples)
s_priors = {'2': s_prior}
obss = bss_cals.BssCals()
s_prior = np.linalg.norm(S[1,:]) * s_prior / np.linalg.norm(s_prior)
Sref = np.stack((S[0,:], s_prior))
Sref = Sref / np.max(Sref,1)[:, None]
S = obss._fix_S(S, s_priors)
np.testing.assert_array_almost_equal(S, Sref, 6,'Wrong H')
def testBoundAFourTes(self):
t2_bounds = {'1':[0, 0.15], '2':[0, 0.3], '3':[0, 0.5], '4':[0, 2]}
t2 = np.array([0.075, 0.15, 0.25, 1])
tes = np.array([0.04, 0.07, 0.13, 0.2])
tes = tes[:, np.newaxis]
f = np.array([0.25, 0.25, 0.25, 0.25])
(_, Aref) = utils.build_A(t2,tes,f)
A = np.ones((4, 4))
obss = bss_cals.BssCals()
A = obss._bound_A(A, t2_bounds, tes)
A = A / np.linalg.norm(A, None, 0)
self.assertIsNotNone(A, 'Missing bounded A')
np.testing.assert_array_almost_equal(A, Aref, 6,'Wrong A')
def testMethodFactorizationMask0(self):
tes = np.array([0.06, 0.1])
tes = tes[:, np.newaxis]
X = np.random.random((2, 30))
mask = 0
params = {'max_iters': 100, 'max_sources': 2 }
r = c = s = 1
result = self.__create_result_object()
obss = bss_cals.BssCals()
obss._method_factorization(X, tes, mask, params, result, r, c, s)
self.assertEqual(np.sum(result.T2[r,c,s,:]),0, 'Unexpected T2 values')
self.assertEqual(np.sum(result.f[r,c,s,:]),0, 'Unexpected f values')
self.assertEqual(result.pd[r,c,s], 0, 'Unexpected S0 values')
self.assertEqual(np.sum(np.sum(result.sources)), 0, 'Unexpected sources')
self.assertEqual(result.rel_error[r,c,s], 0, 'Unexpected S0 values')
@contact: [email protected] @license: LPGL ''' from dwybss.bss import bss_cals import numpy as np import os import shutil # Define paths dpath = os.path.join(os.getcwd(), 'data') opath = os.path.join(os.getcwd(), 'output') shutil.rmtree(opath, True) os.makedirs(opath) # Define data and priors obss = bss_cals.BssCals() max_sources = 2 tes = np.array([0.0751, 0.1351]) tes = tes[:, np.newaxis] dwi = [ os.path.join(dpath, 'data_TE1.nii.gz'), os.path.join(dpath, 'data_TE2.nii.gz') ] data = {'dwi': dwi, 'te': tes} mask = os.path.join(dpath, 'mask.nii.gz') b_values = os.path.join(dpath, 'test.bval') # Run BSS niiout = obss.fwe(data, mask, b_values, opath) print(niiout)