def get(self, structure_graph_id):
# from: https://allensdk.readthedocs.io/en/latest/_static/examples/nb/reference_space.html#Constructing-a-structure-tree
# http://help.brain-map.org/display/api/Atlas+Drawings+and+Ontologies
# ! we need to use different FOLDERS for each structure-graph. else, structures.json gets overwritten by the last retrieved species/graph-id. yikes...
rspc = ReferenceSpaceCache(
self.resolution,
self.reference_space_key,
manifest=f'cache\\reference_space\\{structure_graph_id}\\cache.json'
)
# doc @ http://api.brain-map.org/doc/Structure.html
tree = rspc.get_structure_tree(structure_graph_id=structure_graph_id)
#annotation, meta = rspc.get_annotation_volume()
# https://allensdk.readthedocs.io/en/latest/_static/examples/nb/reference_space.html
name_map = tree.get_name_map()
# from the readthedocs.io-sample: create a mapping from id to acronym.
acronym_map = tree.value_map(
lambda x: x['id'], lambda y: y['acronym']
) # tree.get_id_acronym_map() # tree.value_map(lambda x: x['id'], lambda y: y['acronym'])
# from https://datatofish.com/dictionary-to-dataframe/
structure_map = pd.DataFrame(list(name_map.items()),
columns=[
'structure_id', 'structure_name'
]).set_index(['structure_id'])
# the index (= structure-id) is mapped to an acronym:
structure_map['acronym'] = structure_map.index.map(acronym_map)
# this transforms the ancestor_id_map into a dataframe containing the structure_id and the respective ancestors.
ancestor_map = pd.DataFrame([
{
"structure_id": k,
# ancestors of a structure are obtained as a list, using get_ancestor_id_map()
# but instead of the ancestor's id, we want the name. hence, we lookup using the name_map
"ancestors": [
name_map[id] for id in reversed(v)
] # we reverse the ancestors to always have the root in first position.
# from there, we go into finer structures from left to right. this way, we can have null-values for the root-nodes
} for k, v in tree.get_ancestor_id_map().items()
])
ancestor_map = ancestor_map.set_index(['structure_id'])
# ancestors are stored as a list in one column. but we want separate columns for each ancestor-level:
# https://stackoverflow.com/questions/35491274/pandas-split-column-of-lists-into-multiple-columns
ancestor_map = ancestor_map.ancestors.apply(
pd.Series).add_prefix('level_')
ancestor_map = ancestor_map.where(pd.notnull(ancestor_map), None)
return structure_map.merge(ancestor_map,
left_index=True,
right_index=True,
how="left").sort_index()
reference_space_key="annotation/ccf_2017"
# use the latest version of the CCF
)
# Download
annotated_volume, _ = spacecache.get_annotation_volume()
template_volume, _ = spacecache.get_template_volume()
# Save tiff stacks:
tifffile.imsave(str(save_dir / "reference.tiff"), template_volume)
tifffile.imsave(str(save_dir / "annotated.tiff"), annotated_volume)
# Download structures tree and meshes:
######################################
oapi = OntologiesApi() # ontologies
struct_tree = spacecache.get_structure_tree() # structures tree
# Find id of set of regions with mesh:
select_set = "Structures whose surfaces are represented by a precomputed mesh"
all_sets = pd.DataFrame(oapi.get_structure_sets())
mesh_set_id = all_sets[all_sets.description == select_set].id.values[0]
structs_with_mesh = struct_tree.get_structures_by_set_id([mesh_set_id])
meshes_dir = (save_dir / "meshes") # directory to save meshes into
space = ReferenceSpaceApi()
for s in structs_with_mesh:
name = s["id"]
try:
space.download_structure_mesh(structure_id=s["id"],