# POTENTIAL = pd.DataFrame(fh['POTENTIAL'], columns=ELECTRODES.index)
# for k in ['SIGMA', 'R', 'ALTITUDE', 'AZIMUTH',]:
# POTENTIAL[k] = fh[k]
#
# POTENTIAL['CONDUCTIVITY'] = CONDUCTIVITY.brain
fourSM = FourSphereModel(CONDUCTIVITY, RADIUS, ELECTRODES)
GND_ELECTRODES = ELECTRODES.index[(POTENTIAL[ELECTRODES.index] == 0).all()]
RECORDING_ELECTRODES = ELECTRODES.index[(POTENTIAL[ELECTRODES.index] !=
0).any()]
reconstructors = {
name: kesi.FunctionalKernelFieldReconstructor([
cls(GND_ELECTRODES, ELECTRODES, ROW) if issubclass(
cls, ElectrodeAware) else cls(ROW)
for _, ROW in POTENTIAL[(POTENTIAL.SIGMA < 1) & (
POTENTIAL.ALTITUDE > np.pi / 2 * 80 / 90)].iterrows()
], 'potential', RECORDING_ELECTRODES)
for name, cls in [
('kCSD', GroundedElectrodeAwarePolarGaussianSourceKCSD3D),
('kESI', PolarGaussianSourceFEM),
]
}
gc.collect()
PHI = pd.DataFrame(index=ELECTRODES.index)
loc_dipole = {}
for name, dipole in dipoles.items():
logger.info('Computing potentials for dipole: ' + name)
src_pos = dipole['src_pos']
snk_pos = dipole['snk_pos']
def createReconstructor(self, name, nodes):
return kesi.FunctionalKernelFieldReconstructor(
self.FIELD_COMPONENTS.values(),
name,
nodes)
ELECTRODES.loc[RECORDING_ELECTRODES], SOURCES_UNIFORM),
(ElectrodeAwareCartesianGaussianSourceKCSD3D, UNIFORM_ELECTRODES,
SOURCES_UNIFORM),
(ElectrodeAwarePolarGaussianSourceKCSD3D,
ELECTRODES.loc[RECORDING_ELECTRODES], SOURCES_GM),
(PolarGaussianSourceFEM, ELECTRODES.loc[RECORDING_ELECTRODES], SOURCES_GM),
]
reconstructors = []
DF = []
for setup, (cls, electrodes, sources) in enumerate(SETUPS):
print(cls.__name__)
reconstructor = kesi.FunctionalKernelFieldReconstructor([
cls(electrodes, ROW) if issubclass(cls, ElectrodeAware) else cls(ROW)
for _, ROW in sources.iterrows()
], 'potential', electrodes.index)
reconstructors.append(reconstructor)
gc.collect()
A = np.array([
f.csd(0, YY.flatten(), ZZ.flatten())
for f in reconstructor._field_components
],
dtype=np.float128).T
B = reconstructor._pre_cross_kernel.astype(np.float128)
N_ELS = reconstructor._pre_cross_kernel.shape[1]
measures_names = ['electrode{:03d}'.format(i) for i in range(N_ELS)]
measures = [np.identity(N_ELS)]
if cls is GaussianSourceFEM: