# -*- coding: utf-8 -*-
#
#
# TheVirtualBrain-Framework Package. This package holds all Data Management, and
# Web-UI helpful to run brain-simulations. To use it, you also need to download
# TheVirtualBrain-Scientific Package (for simulators). See content of the
# documentation-folder for more details. See also http://www.thevirtualbrain.org
#
# (c) 2012-2024, Baycrest Centre for Geriatric Care ("Baycrest") and others
#
# This program is free software: you can redistribute it and/or modify it under the
# terms of the GNU General Public License as published by the Free Software Foundation,
# either version 3 of the License, or (at your option) any later version.
# This program is distributed in the hope that it will be useful, but WITHOUT ANY
# WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A
# PARTICULAR PURPOSE. See the GNU General Public License for more details.
# You should have received a copy of the GNU General Public License along with this
# program. If not, see <http://www.gnu.org/licenses/>.
#
#
# CITATION:
# When using The Virtual Brain for scientific publications, please cite it as explained here:
# https://www.thevirtualbrain.org/tvb/zwei/neuroscience-publications
#
#
"""
Backend-side for Visualizers that display measures on regions in the brain volume.
.. moduleauthor:: Andrei Mihai <mihai.andrei@codemart.ro>
"""
import json
from abc import ABCMeta
from six import add_metaclass
from tvb.adapters.datatypes.db.graph import ConnectivityMeasureIndex
from tvb.adapters.datatypes.db.region_mapping import RegionVolumeMappingIndex
from tvb.adapters.datatypes.db.structural import StructuralMRIIndex
from tvb.adapters.datatypes.db.volume import VolumeIndex
from tvb.adapters.datatypes.h5.time_series_h5 import TimeSeriesRegionH5
from tvb.basic.neotraits.api import Attr
from tvb.core.adapters.abcadapter import ABCAdapterForm
from tvb.core.adapters.abcdisplayer import ABCDisplayer, URLGenerator
from tvb.core.adapters.arguments_serialisation import *
from tvb.core.adapters.exceptions import LaunchException
from tvb.core.entities.filters.chain import FilterChain
from tvb.core.entities.model.model_datatype import DataTypeMatrix
from tvb.core.entities.storage import dao
from tvb.core.neocom import h5
from tvb.core.neotraits.forms import TraitDataTypeSelectField, StrField
from tvb.core.neotraits.view_model import ViewModel, DataTypeGidAttr
from tvb.datatypes.graph import ConnectivityMeasure
from tvb.datatypes.region_mapping import RegionVolumeMapping
from tvb.datatypes.structural import StructuralMRI
@add_metaclass(ABCMeta)
class _MappedArrayVolumeBase(ABCDisplayer):
"""
Base functionality for all non-temporal volume views.
It prepares for display a slice of a mapped array.
"""
_ui_subsection = "volume"
def get_required_memory_size(self, view_model):
# type: (BaseVolumeVisualizerModel) -> int
return -1
@staticmethod
def get_default_slice(measure_shape, nregions):
default = [0 for _ in range(len(measure_shape))]
for i in range(len(measure_shape)):
if measure_shape[i] == nregions:
default[i] = slice(None)
return tuple(default)
raise LaunchException("The mapped array of shape %s is incompatible with the region mapping "
"(expected values for %d connectivity regions)." % (measure_shape, nregions))
def _ensure_region_mapping_index(self, region_mapping_volume, measure=None):
# type: (RegionVolumeMappingIndex, DataTypeMatrix) -> RegionVolumeMappingIndex
if region_mapping_volume is None:
if measure is not None and hasattr(measure, "fk_connectivity_gid"):
region_mapping_volume = dao.get_generic_entity(RegionVolumeMappingIndex,
measure.fk_connectivity_gid, 'fk_connectivity_gid')
if region_mapping_volume is not None and len(region_mapping_volume):
region_mapping_volume = region_mapping_volume[0]
else:
region_mapping_volume = None
else:
region_mapping_volume = dao.try_load_last_entity_of_type(self.current_project_id,
RegionVolumeMappingIndex)
if region_mapping_volume is None:
raise LaunchException('You should have a compatible volume mapping to launch this viewer')
return region_mapping_volume
def _compute_region_volume_map_params(self, region_mapping_volume):
# type: (RegionVolumeMappingIndex) -> dict
# prepare the url that will display the region volume map
conn_index = dao.get_datatype_by_gid(region_mapping_volume.fk_connectivity_gid)
min_value, max_value = [0, conn_index.number_of_regions]
url_volume_data = URLGenerator.build_url(self.stored_adapter.id, 'get_volume_view',
region_mapping_volume.gid, '')
return dict(minValue=min_value, maxValue=max_value, urlVolumeData=url_volume_data)
def _compute_measure_params(self, rvm_index, measure, data_slice):
# type: (RegionVolumeMappingIndex, DataTypeMatrix, str) -> dict
# prepare the url that will project the measure onto the region volume map
measure_shape = measure.parsed_shape
if not data_slice:
conn_index = dao.get_datatype_by_gid(rvm_index.fk_connectivity_gid)
data_slice = self.get_default_slice(measure_shape, conn_index.number_of_regions)
data_slice = slice_str(data_slice)
url_volume_data = URLGenerator.build_url(self.stored_adapter.id, 'get_mapped_array_volume_view',
rvm_index.gid, parameter='')
url_volume_data += 'mapped_array_gid=' + measure.gid + ';mapped_array_slice=' + data_slice + ';'
return dict(minValue=measure.array_data_min, maxValue=measure.array_data_max,
urlVolumeData=url_volume_data, measureShape=measure.shape, measureSlice=data_slice)
def get_mapped_array_volume_view(self, entity_gid, mapped_array_gid, x_plane, y_plane, z_plane,
mapped_array_slice=None, **kwargs):
with h5.h5_file_for_gid(entity_gid) as entity_h5:
data_shape = entity_h5.array_data.shape
x_plane, y_plane, z_plane = preprocess_space_parameters(x_plane, y_plane, z_plane, data_shape[0],
data_shape[1], data_shape[2])
slice_x, slice_y, slice_z = entity_h5.get_volume_slice(x_plane, y_plane, z_plane)
connectivity_gid = entity_h5.connectivity.load()
with h5.h5_file_for_gid(mapped_array_gid) as mapped_array_h5:
if mapped_array_slice:
matrix_slice = parse_slice(mapped_array_slice)
measure = mapped_array_h5.array_data[matrix_slice]
else:
measure = mapped_array_h5.array_data[:]
connectivity_index = self.load_entity_by_gid(connectivity_gid)
if measure.shape != (connectivity_index.number_of_regions,):
raise ValueError('cannot project measure on the space')
result_x = measure[slice_x]
result_y = measure[slice_y]
result_z = measure[slice_z]
# Voxels outside the brain are -1. The indexing above is incorrect for those voxels as it
# associates the values of the last region measure[-1] to them.
# Here we replace those values with an out of scale value.
result_x[slice_x == -1] = measure.min() - 1
result_y[slice_y == -1] = measure.min() - 1
result_z[slice_z == -1] = measure.min() - 1
return [[result_x.tolist()],
[result_y.tolist()],
[result_z.tolist()]]
@staticmethod
def compute_background_params(min_value=0, max_value=0, url=None):
return dict(minBackgroundValue=min_value, maxBackgroundValue=max_value, urlBackgroundVolumeData=url)
def get_voxel_region(self, region_mapping_volume_gid, x_plane, y_plane, z_plane):
with h5.h5_file_for_gid(region_mapping_volume_gid) as entity_h5:
data_shape = entity_h5.array_data.shape
x_plane, y_plane, z_plane = preprocess_space_parameters(x_plane, y_plane, z_plane, data_shape[0],
data_shape[1], data_shape[2])
slices = slice(x_plane, x_plane + 1), slice(y_plane, y_plane + 1), slice(z_plane, z_plane + 1)
voxel = entity_h5.array_data[slices][0, 0, 0]
connectivity_gid = entity_h5.connectivity.load()
if voxel != -1:
conn = self.load_traited_by_gid(connectivity_gid)
return conn.region_labels[int(voxel)]
else:
return 'background'
def compute_params(self, region_mapping_volume=None, measure=None, data_slice='', background=None):
# type: (RegionVolumeMappingIndex, DataTypeMatrix, str, StructuralMRIIndex) -> dict
region_mapping_volume = self._ensure_region_mapping_index(region_mapping_volume, measure)
url_voxel_region = URLGenerator.build_url(self.stored_adapter.id, 'get_voxel_region', region_mapping_volume.gid,
parameter='')
if measure is None:
params = self._compute_region_volume_map_params(region_mapping_volume)
else:
params = self._compute_measure_params(region_mapping_volume, measure, data_slice)
volume_gid = region_mapping_volume.fk_volume_gid
volume_index = self.load_entity_by_gid(volume_gid)
assert isinstance(volume_index, VolumeIndex)
volume_shape = region_mapping_volume.parsed_shape
volume_shape = (1,) + volume_shape
params.update(volumeShape=json.dumps(volume_shape),
volumeOrigin=volume_index.origin,
voxelUnit=volume_index.voxel_unit,
voxelSize=volume_index.voxel_size,
urlVoxelRegion=url_voxel_region)
if background is None:
background = dao.try_load_last_entity_of_type(self.current_project_id, StructuralMRIIndex)
if background is None: # still
params.update(self.compute_background_params())
else:
url_volume_data = URLGenerator.build_url(self.stored_adapter.id, 'get_volume_view', background.gid, '')
params.update(self.compute_background_params(background.array_data_min,
background.array_data_max, url_volume_data))
return params
def get_volume_view(self, entity_gid, **kwargs):
with h5.h5_file_for_gid(entity_gid) as ts_region_h5:
if isinstance(ts_region_h5, TimeSeriesRegionH5):
return self.prepare_view_region(ts_region_h5, **kwargs)
volume_view = ts_region_h5.get_volume_view(**kwargs)
return volume_view
def prepare_view_region(self, ts_h5, x_plane, y_plane, z_plane, from_idx=None, to_idx=None, var=0, mode=0):
region_mapping_volume_gid = ts_h5.region_mapping_volume.load()
if region_mapping_volume_gid is None:
raise Exception("Invalid method called for TS without Volume Mapping!")
with h5.h5_file_for_gid(region_mapping_volume_gid) as volume_rm_h5:
volume_rm_shape = volume_rm_h5.array_data.shape
# Work with space inside Volume:
x_plane, y_plane, z_plane = preprocess_space_parameters(x_plane, y_plane, z_plane, volume_rm_shape[0],
volume_rm_shape[1], volume_rm_shape[2])
region = self.get_view_region(ts_h5, volume_rm_h5, from_idx, to_idx, x_plane, y_plane, z_plane, var, mode)
return region
def get_view_region(self, ts_h5, volume_rm_h5, from_idx, to_idx, x_plane, y_plane, z_plane, var=0, mode=0):
"""
Retrieve 3 slices through the Volume TS, at the given X, y and Z coordinates, and in time [from_idx .. to_idx].
:param ts_h5: input TimeSeriesH5
:param volume_rm_h5: input RegionVolumeMappingH5
:param from_idx: int This will be the limit on the first dimension (time)
:param to_idx: int Also limit on the first Dimension (time)
:param x_plane: int coordinate
:param y_plane: int coordinate
:param z_plane: int coordinate
:return: An array of 3 Matrices 2D, each containing the values to display in planes xy, yz and xy.
"""
var, mode = int(var), int(mode)
slice_x, slice_y, slice_z = volume_rm_h5.get_volume_slice(x_plane, y_plane, z_plane)
# Read from the current TS:
from_idx, to_idx, current_time_length = preprocess_time_parameters(from_idx, to_idx, ts_h5.data.shape[0])
no_of_regions = ts_h5.data.shape[2]
time_slices = slice(from_idx, to_idx), slice(var, var + 1), slice(no_of_regions), slice(mode, mode + 1)
min_signal = ts_h5.get_min_max_values()[0]
regions_ts = ts_h5.read_data_slice(time_slices)[:, 0, :, 0]
regions_ts = numpy.hstack((regions_ts, numpy.ones((current_time_length, 1)) * ts_h5.out_of_range(min_signal)))
# Index from TS with the space mapping:
result_x, result_y, result_z = [], [], []
for i in range(0, current_time_length):
result_x.append(regions_ts[i][slice_x].tolist())
result_y.append(regions_ts[i][slice_y].tolist())
result_z.append(regions_ts[i][slice_z].tolist())
return [result_x, result_y, result_z]
[docs]
class BaseVolumeVisualizerModel(ViewModel):
background = DataTypeGidAttr(
linked_datatype=StructuralMRI,
required=False,
label='Background T1'
)
[docs]
class VolumeVisualizerModel(BaseVolumeVisualizerModel):
measure = DataTypeGidAttr(
linked_datatype=DataTypeMatrix,
label='Measure',
doc='A measure to view on anatomy'
)
region_mapping_volume = DataTypeGidAttr(
linked_datatype=RegionVolumeMapping,
required=False,
label='Region mapping'
)
data_slice = Attr(
field_type=str,
required=False,
label='slice indices in numpy syntax'
)
[docs]
class MappedArrayVolumeVisualizer(_MappedArrayVolumeBase):
"""
This is a generic mapped array visualizer on a region volume.
To view a multidimensional array one has to give this viewer a slice.
"""
_ui_name = "Array Volume Visualizer"
[docs]
def launch(self, view_model):
# type: (VolumeVisualizerModel) -> dict
measure_index = self.load_entity_by_gid(view_model.measure)
region_mapping_volume_index = None
background_volume_index = None
if view_model.region_mapping_volume:
region_mapping_volume_index = self.load_entity_by_gid(view_model.region_mapping_volume)
if view_model.background:
background_volume_index = self.load_entity_by_gid(view_model.background)
params = self.compute_params(region_mapping_volume_index, measure_index, view_model.data_slice,
background=background_volume_index)
params['title'] = "Mapped array on region volume Visualizer"
return self.build_display_result("time_series_volume/staticView", params,
pages=dict(controlPage="time_series_volume/controls"))
[docs]
class ConnectivityMeasureVolumeVisualizerModel(BaseVolumeVisualizerModel):
connectivity_measure = DataTypeGidAttr(
linked_datatype=ConnectivityMeasure,
label='Connectivity measure',
doc='A connectivity measure'
)
region_mapping_volume = DataTypeGidAttr(
linked_datatype=RegionVolumeMapping,
required=False,
label='Region mapping'
)
[docs]
class ConnectivityMeasureVolumeVisualizer(_MappedArrayVolumeBase):
_ui_name = "Connectivity Measure Volume Visualizer"
[docs]
def launch(self, view_model):
# type: (ConnectivityMeasureVolumeVisualizerModel) -> dict
connectivity_measure_index = self.load_entity_by_gid(view_model.connectivity_measure)
region_mapping_volume_index = None
background_volume_index = None
if view_model.region_mapping_volume:
region_mapping_volume_index = self.load_entity_by_gid(view_model.region_mapping_volume)
if view_model.background:
background_volume_index = self.load_entity_by_gid(view_model.background)
params = self.compute_params(region_mapping_volume_index, connectivity_measure_index,
background=background_volume_index)
params['title'] = "Connectivity Measure in Volume Visualizer"
# the view will display slicing information if this key is present.
# compute_params works with generic mapped arrays and it will return slicing info
del params['measureSlice']
return self.build_display_result("time_series_volume/staticView", params,
pages=dict(controlPage="time_series_volume/controls"))
[docs]
class RegionVolumeMappingVisualiserModel(BaseVolumeVisualizerModel):
region_mapping_volume = DataTypeGidAttr(
linked_datatype=RegionVolumeMapping,
label='Region mapping'
)
connectivity_measure = DataTypeGidAttr(
linked_datatype=ConnectivityMeasure,
required=False,
label='Connectivity measure',
doc='A connectivity measure'
)
[docs]
class RegionVolumeMappingVisualiser(_MappedArrayVolumeBase):
_ui_name = "Region Volume Mapping Visualizer"
[docs]
def launch(self, view_model):
# type: (RegionVolumeMappingVisualiserModel) -> dict
connectivity_measure_index = None
region_mapping_volume_index = None
background_volume_index = None
if view_model.connectivity_measure:
connectivity_measure_index = self.load_entity_by_gid(view_model.connectivity_measure)
if view_model.region_mapping_volume:
region_mapping_volume_index = self.load_entity_by_gid(view_model.region_mapping_volume)
if view_model.background:
background_volume_index = self.load_entity_by_gid(view_model.background)
params = self.compute_params(region_mapping_volume_index, connectivity_measure_index,
background=background_volume_index)
params['title'] = "Volume to Regions Visualizer"
return self.build_display_result("time_series_volume/staticView", params,
pages=dict(controlPage="time_series_volume/controls"))
[docs]
class MriVolumeVisualizer(_MappedArrayVolumeBase):
_ui_name = "MRI Volume Visualizer"
_ui_subsection = "volume"
[docs]
def get_required_memory_size(self, view_model):
# type: (BaseVolumeVisualizerModel) -> int
return -1
[docs]
def launch(self, view_model):
# type: (BaseVolumeVisualizerModel) -> dict
structural_mri = self.load_entity_by_gid(view_model.background)
assert isinstance(structural_mri, StructuralMRIIndex)
volume_shape = structural_mri.parsed_shape
volume_shape = (1,) + volume_shape
url_volume_data = URLGenerator.build_url(self.stored_adapter.id, 'get_volume_view', view_model.background, '')
volume_gid = structural_mri.fk_volume_gid
volume_index = self.load_entity_by_gid(volume_gid)
assert isinstance(volume_index, VolumeIndex)
params = dict(title="MRI Volume visualizer",
minValue=structural_mri.array_data_min, maxValue=structural_mri.array_data_max,
urlVolumeData=url_volume_data, volumeShape=json.dumps(volume_shape),
volumeOrigin=volume_index.origin, voxelUnit=volume_index.voxel_unit,
voxelSize=volume_index.voxel_size,
urlVoxelRegion='', urlBackgroundVolumeData='',
minBackgroundValue=structural_mri.array_data_min,
maxBackgroundValue=structural_mri.array_data_max)
return self.build_display_result("time_series_volume/staticView", params,
pages=dict(controlPage="time_series_volume/controls"))