def test_phase_correct_zero():
"""
Test that phase correction works
"""
# Make some complex numbers, with the same phase (pi):
arr = np.random.rand(10,10,10,1024) * np.exp(1j * np.pi)
corr_arr = ut.phase_correct_zero(arr, -np.pi)
npt.assert_array_almost_equal(np.angle(corr_arr), 0)
def on_show(self):
self.axes.clear()
self.axes.set_xlabel('ppm')
self.axes.grid(True)
has_series = False
self.data.get_spectra(self.line_spin.value(), self.cutoff_spin.value())
for row in range(self.series_list_model.rowCount()):
model_index = self.series_list_model.index(row, 0)
checked = self.series_list_model.data(
model_index, QtCore.Qt.CheckStateRole) == QtCore.Qt.Checked
name = str(self.series_list_model.data(model_index))
if checked:
has_series = True
phase_zero = self.phase_spin.value()
series = self.data.get_series_data(name)
series = ut.phase_correct_zero(series, phase_zero)
self.axes.plot(self.data.f_ppm, series, label=name)
if has_series and self.legend_cb.isChecked():
self.axes.legend()
self.canvas.draw()
def fit_gaba(self,
reject_outliers=3.0,
fit_lb=2.8,
fit_ub=3.4,
phase_correct=True,
fit_func=None):
"""
Fit either a single Gaussian, or a two-Gaussian to the GABA 3 PPM
peak.
Parameters
----------
reject_outliers : float
Z-score criterion for rejection of outliers, based on their model
parameter
fit_lb, fit_ub : float
Frequency bounds (in ppm) for the region of the spectrum to be
fit.
phase_correct : bool
Where to perform zero-order phase correction based on the fit of
the creatine peaks in the sum spectra
fit_func : None or callable (default None).
If this is set to `False`, an automatic selection will take place,
choosing between a two-Gaussian and a single Gaussian, based on a
split-half cross-validation procedure. Otherwise, the requested
callable function will be fit. Needs to conform to the conventions
of `fit_gaussian`/`fit_two_gaussian` and
`ut.gaussian`/`ut.two_gaussian`.
"""
# We need to fit the creatine, so that we know which transients to
# exclude in fitting this peak:
if not hasattr(self, 'creatine_params'):
self.fit_creatine()
fit_spectra = np.ones(self.diff_spectra.shape) * np.nan
# Silence warnings:
with warnings.catch_warnings():
warnings.simplefilter("ignore")
fit_spectra =\
self.diff_spectra[self._cr_transients].copy()
if phase_correct:
for ii, this_spec in enumerate(fit_spectra):
# Silence warnings:
with warnings.catch_warnings():
warnings.simplefilter("ignore")
fit_spectra[ii] = ut.phase_correct_zero(
this_spec,
self.creatine_params[self._cr_transients][ii, 3])
if fit_func is None:
# Cross-validate!
fitter, self.gaba_func, self.gaba_model_err, self.gaba_signal_err=\
self._xval_choose_funcs(fit_spectra,
reject_outliers,
fit_lb, fit_ub)
# Otherwise, you had better supply a couple of callables that can be
# used to fit these spectra!
else:
fitter = fit_func[0]
self.gaba_func = fit_func[1]
self.gaba_model_err, self.gaba_signal_err = \
self._xval_model_error(fit_spectra, reject_outliers,
fit_lb, fit_ub, fitter, self.gaba_func)
# Either way, we end up fitting to everything in the end:
choose_transients, model, signal, params, this_idx = self._fit_helper(
fit_spectra, reject_outliers, fit_lb, fit_ub, fitter)
self._gaba_transients = choose_transients
self.gaba_model = model
self.gaba_signal = signal
self.gaba_params = params
self.gaba_idx = this_idx
mean_params = stats.nanmean(params, 0)
self.gaba_auc = self._calc_auc(self.gaba_func, params, self.gaba_idx)
def fit_gaba(self, reject_outliers=3.0, fit_lb=2.8, fit_ub=3.4,
phase_correct=True, fit_func=None):
"""
Fit either a single Gaussian, or a two-Gaussian to the GABA 3 PPM
peak.
Parameters
----------
reject_outliers : float
Z-score criterion for rejection of outliers, based on their model
parameter
fit_lb, fit_ub : float
Frequency bounds (in ppm) for the region of the spectrum to be
fit.
phase_correct : bool
Where to perform zero-order phase correction based on the fit of
the creatine peaks in the sum spectra
fit_func : None or callable (default None).
If this is set to `False`, an automatic selection will take place,
choosing between a two-Gaussian and a single Gaussian, based on a
split-half cross-validation procedure. Otherwise, the requested
callable function will be fit. Needs to conform to the conventions
of `fit_gaussian`/`fit_two_gaussian` and
`ut.gaussian`/`ut.two_gaussian`.
"""
# We need to fit the creatine, so that we know which transients to
# exclude in fitting this peak:
if not hasattr(self, 'creatine_params'):
self.fit_creatine()
fit_spectra = np.ones(self.diff_spectra.shape) * np.nan
# Silence warnings:
with warnings.catch_warnings():
warnings.simplefilter("ignore")
fit_spectra =\
self.diff_spectra[self._cr_transients].copy()
if phase_correct:
for ii, this_spec in enumerate(fit_spectra):
# Silence warnings:
with warnings.catch_warnings():
warnings.simplefilter("ignore")
fit_spectra[ii] = ut.phase_correct_zero(this_spec,
self.creatine_params[self._cr_transients][ii, 3])
if fit_func is None:
# Cross-validate!
fitter, self.gaba_func, self.gaba_model_err, self.gaba_signal_err=\
self._xval_choose_funcs(fit_spectra,
reject_outliers,
fit_lb, fit_ub)
# Otherwise, you had better supply a couple of callables that can be
# used to fit these spectra!
else:
fitter = fit_func[0]
self.gaba_func = fit_func[1]
self.gaba_model_err, self.gaba_signal_err = \
self._xval_model_error(fit_spectra, reject_outliers,
fit_lb, fit_ub, fitter, self.gaba_func)
# Either way, we end up fitting to everything in the end:
choose_transients, model, signal, params, this_idx = self._fit_helper(
fit_spectra, reject_outliers,
fit_lb, fit_ub, fitter)
self._gaba_transients = choose_transients
self.gaba_model = model
self.gaba_signal = signal
self.gaba_params = params
self.gaba_idx = this_idx
mean_params = stats.nanmean(params, 0)
self.gaba_auc = self._calc_auc(self.gaba_func, params, self.gaba_idx)