def __init__(
self,
base_dir=None,
add_root=True,
use_cache=True,
scene_kwargs={},
**kwargs,
):
"""
Initialise the class instance to get a few useful paths and variables.
:param base_dir: str, path to base directory in which all of brainrender data are stored.
Pass only if you want to use a different one from what's default.
:param add_root: bool, if True the root mesh is added to the rendered scene
:param use_cache: if true data are loaded from a cache to speed things up.
Useful to set it to false to help debugging.
:param scene_kwargs: dict, params passed to the instance of Scene associated with this class
"""
Paths.__init__(self, base_dir=base_dir, **kwargs)
# Get MCM cache
cache_path = (Path(self.mouse_connectivity_volumetric) /
"voxel_model_manifest.json")
if not cache_path.exists():
if not connected_to_internet():
raise ValueError(
"The first time you use this class it will need to download some data, but it seems that you're not connected to the internet."
)
print(
"Downloading volumetric data. This will take several minutes but it only needs to be done once."
)
self.cache = VoxelModelCache(manifest_file=str(cache_path))
self.voxel_array = None
self.target_coords, self.source_coords = None, None
# Get projection cache paths
self.data_cache = self.mouse_connectivity_volumetric_cache
self.data_cache_projections = os.path.join(self.data_cache,
"projections")
self.data_cache_targets = os.path.join(self.data_cache, "targets")
self.data_cache_sources = os.path.join(self.data_cache, "sources")
for fold in [
self.data_cache_projections,
self.data_cache_targets,
self.data_cache_sources,
]:
if not os.path.isdir(fold):
os.mkdir(fold)
# Get structures tree
self.structure_tree = self.cache.get_structure_tree()
# Get scene
self.scene = Scene(add_root=add_root, **scene_kwargs)
# Other vars
self.use_cache = use_cache
def main():
input_data = ju.read(INPUT_JSON)
manifest_file = input_data.get('manifest_file')
manifest_file = os.path.join(TOP_DIR, manifest_file)
log_level = input_data.get('log_level', logging.DEBUG)
logging.getLogger().setLevel(log_level)
# get cache, metric
logging.debug("loading regional matrix")
cache = VoxelModelCache(manifest_file=manifest_file)
df_metric = cache.get_normalized_connection_density(dataframe=True)
# plot
fig = plot(df_metric,
STRUCTURES,
cache,
GRID_KWS,
CBAR_KWS,
HEATMAP_KWS,
figsize=FIGSIZE)
fig.savefig(OUTPUT_FILE, **SAVEFIG_KWARGS)
plt.close(fig)
def voxel_model_cache(fn_temp_dir, mcc):
manifest_path = os.path.join(fn_temp_dir, 'voxel_model_manifest.json')
cache = VoxelModelCache(manifest_file=manifest_path)
cache.get_reference_space = mock.Mock()
cache.get_reference_space.return_value = mcc.get_reference_space()
return cache
def main():
input_data = ju.read(INPUT_JSON)
manifest_file = input_data.get('manifest_file')
manifest_file = os.path.join(TOP_DIR, manifest_file)
log_level = input_data.get('log_level', logging.DEBUG)
logging.getLogger().setLevel(log_level)
# get cache, metric
logging.debug("loading regional matrix")
cache = VoxelModelCache(manifest_file=manifest_file)
df_metric = cache.get_normalized_connection_density(dataframe=True)
logging.debug("getting cortical network")
df_cortex = get_cortical_df(df_metric, cache)
# get projection types
full_ipsi, cortex_ipsi = get_pt((df_metric, df_cortex))
full_contra, cortex_contra = get_pt((df_metric, df_cortex), pt="contra")
logging.debug("Computing gaussian mixture model fits for max: %s" %
MAX_COMPONENTS)
dfs = (full_ipsi, full_contra, cortex_ipsi, cortex_contra)
labels = ("full-ipsi", "full-contra", "cortex-ipsi", "cortex-contra")
frames = []
for d, l in zip(dfs, labels):
# log transform
d = np.log10(d[d > 0]).reshape(-1, 1)
# normality test
_, p_value = stats.shapiro(d)
# gmm
gmm, bic = fit_gmm(d, MAX_COMPONENTS, **GMM_PARAMS)
columns = ('mean', 'var', 'weight')
print("", l, "-" * 40, sep="\n")
print("shapiro-wilk p_value : %.5g" % p_value)
print("optimal n components : %d" % gmm.n_components)
print("bic : %.5g" % bic)
print('\t'.join(columns))
print("----\t---\t------")
attrs = tuple(
map(np.ravel, (gmm.means_, gmm.covariances_, gmm.weights_)))
for x in zip(*attrs):
print("%.2f\t%.2f\t%.3f" % x)
df = pd.DataFrame(dict(zip(columns, attrs)))
df.index.name = 'n_components'
frames.append(df)
df = pd.concat(frames, keys=labels).unstack()
df.to_csv(OUTPUT_FILE)
def test_to_json(fn_temp_dir):
# ------------------------------------------------------------------------
# tests JSON serialization
manifest_file = 'manifest.json'
resolution = 100
path = os.path.join(fn_temp_dir, 'output.json')
cache = VoxelModelCache(manifest_file=manifest_file, resolution=resolution)
cache.to_json(path)
input_data = json_utilities.read(path)
assert input_data['manifest_file'] == manifest_file
assert input_data['resolution'] == resolution
def main(injection_region, filtered=False):
input_data = ju.read(INPUT_JSON)
manifest_file = input_data.get('manifest_file')
manifest_file = os.path.join(TOP_DIR, manifest_file)
log_level = input_data.get('log_level', logging.DEBUG)
logging.getLogger().setLevel(log_level)
# get voxel_model_cache
cache = VoxelModelCache(manifest_file=manifest_file)
# get region id
region_id = get_region_id(cache, injection_region)
logging.debug("performing virtual injection into %s (%s)" %
(injection_region, region_id))
projection = get_projection(
cache, region_id, full=FULL_INJECTION, filtered=filtered)
# get projection (row)
logging.debug("upscaling projection to 10 micron")
projection = upscale_projection(projection, SCALE, **UPSCALE_KWARGS)
# file name
suffix = injection_region + "full" if FULL_INJECTION else injection_region
vol_file = os.path.join(VOLUME_DIR, "projection_density_%s.nrrd" % suffix)
logging.debug("saving projection volume : %s" % vol_file)
nrrd.write(vol_file, projection, options=dict(encoding='raw'))
return vol_file
def main():
input_data = ju.read(INPUT_JSON)
structures = input_data.get('structures')
manifest_file = input_data.get('manifest_file')
manifest_file = os.path.join(TOP_DIR, manifest_file)
log_level = input_data.get('log_level', logging.DEBUG)
logging.getLogger().setLevel(log_level)
# experiments to exclude
experiments_exclude = ju.read(EXPERIMENTS_EXCLUDE_JSON)
# get caching object
cache = VoxelModelCache(manifest_file=manifest_file)
output_file = os.path.join(OUTPUT_DIR, 'hyperparameters-%s.json' % OPTION)
results = dict()
for structure in structures:
logging.debug("Running cross validation for structure: %s", structure)
structure_id = get_structure_id(cache, structure)
results[structure] = fit_structure(cache,
structure_id,
experiments_exclude,
kernel=KERNEL,
model_option=OPTION)
# write results
ju.write(output_file, results)
def main():
input_data = ju.read(INPUT_JSON)
manifest_file = input_data.get('manifest_file')
manifest_file = os.path.join(TOP_DIR, manifest_file)
log_level = input_data.get('log_level', logging.DEBUG)
logging.getLogger().setLevel(log_level)
# experiments to exclude
experiments_exclude = ju.read(EXPERIMENTS_EXCLUDE_JSON)
# load hyperparameter dict
suffix = 'high_res' if HIGH_RES else 'standard'
# get caching object
cache = VoxelModelCache(manifest_file=manifest_file)
fit_kwargs = dict(high_res=HIGH_RES,
threshold_injection=THRESHOLD_INJECTION,
experiments_exclude=experiments_exclude)
model = fit(cache, **fit_kwargs)
# write results
logging.debug('saving')
output_file = os.path.join(OUTPUT_DIR, 'homogeneous-%s-model.csv' % suffix)
model.to_csv(output_file)
def main():
input_data = ju.read(INPUT_JSON)
structures = input_data.get('structures')
manifest_file = input_data.get('manifest_file')
manifest_file = os.path.join(TOP_DIR, manifest_file)
log_level = input_data.get('log_level', logging.DEBUG)
logging.getLogger().setLevel(log_level)
# experiments to exclude
experiments_exclude = ju.read(EXPERIMENTS_EXCLUDE_JSON)
# get caching object
cache = VoxelModelCache(manifest_file=manifest_file)
# get full map
full_map = get_full_map(cache, structures, experiments_exclude)
# convert to df
df = get_df(full_map, structures)
# save
df.to_csv(OUTPUT_FILE)
def main():
# set log level
logging.getLogger().setLevel(args.log_level)
# initialize cache object
logging.info('initializing VoxelModelCache with manifest_file: %s',
args.manifest_file)
cache = VoxelModelCache(manifest_file=args.manifest_file)
structure_ids = get_ordered_summary_structures(cache)
# load in voxel model
logging.info('loading array')
voxel_array, source_mask, target_mask = cache.get_voxel_connectivity_array()
source_key = source_mask.get_key(structure_ids=structure_ids)
ipsi_key = target_mask.get_key(structure_ids=structure_ids, hemisphere_id=2)
contra_key = target_mask.get_key(structure_ids=structure_ids, hemisphere_id=1)
ipsi_model = RegionalizedModel.from_voxel_array(
voxel_array, source_key, ipsi_key, ordering=structure_ids, dataframe=True)
contra_model = RegionalizedModel.from_voxel_array(
voxel_array, source_key, contra_key, ordering=structure_ids, dataframe=True)
# get each metric
get_metric = lambda s: pd.concat((getattr(ipsi_model, s), getattr(contra_model, s)),
keys=('ipsi', 'contra'), axis=1)
# write results
if not os.path.exists(OUTPUT_DIR):
os.makedirs(OUTPUT_DIR)
# regionalized
logging.info('saving metrics to directory: %s', OUTPUT_DIR)
get_metric('connection_density').to_csv(
os.path.join(OUTPUT_DIR, 'connection_density.csv'))
get_metric('connection_strength').to_csv(
os.path.join(OUTPUT_DIR, 'connection_strength.csv'))
get_metric('normalized_connection_density').to_csv(
os.path.join(OUTPUT_DIR, 'normalized_connection_density.csv'))
get_metric('normalized_connection_strength').to_csv(
os.path.join(OUTPUT_DIR, 'normalized_connection_strength.csv'))
def main():
# get cache object for loading annotation/model objects
cache = VoxelModelCache(manifest_file=MANIFEST_FILE)
# get ids for visual areas
visual_ids = get_structure_ids_from_acronyms(cache, VISUAL_AREA_ACRONYMS)
# get voxel-scale connectivity of visual network (ipsilateral)
visual_network = get_voxel_subgraph(cache, visual_ids, hemisphere_id=2)
# save visual_network
logger.debug('Saving the visual network to %s', OUTPUT_FILE)
np.savetxt(OUTPUT_FILE, visual_network, delimiter='')
def main():
input_data = ju.read(INPUT_JSON)
manifest_file = input_data.get('manifest_file')
manifest_file = os.path.join(TOP_DIR, manifest_file)
log_level = input_data.get('log_level', logging.DEBUG)
logging.getLogger().setLevel(log_level)
# get cache, metric
logging.debug("loading regional matrix")
cache = VoxelModelCache(manifest_file=manifest_file)
df_metric = cache.get_normalized_connection_density(dataframe=True)
logging.debug("getting cortical network")
df_cortex = get_cortical_df(df_metric, cache)
# get projection types
full_ipsi, cortex_ipsi = get_pt((df_metric, df_cortex))
full_contra, cortex_contra = get_pt((df_metric, df_cortex), pt="contra")
logging.debug("Computing distribution fits for")
logging.debug("%s" % DISTRIBUTIONS)
fitter = DistFit(DISTRIBUTIONS)
dfs = (full_ipsi, full_contra, cortex_ipsi, cortex_contra)
labels = ("full-ipsi", "full-contra", "cortex-ipsi", "cortex-contra")
frames = []
for d, l in zip(dfs, labels):
fitter.fit(d[d > 0])
logging.debug(l)
logging.debug(str(fitter))
frames.append(results_to_df(fitter))
df = pd.concat(frames, keys=labels).unstack()
df.to_csv(OUTPUT_FILE)
def test_from_json(fn_temp_dir):
# ------------------------------------------------------------------------
# tests alternative constructor
manifest_file = 'manifest.json'
resolution = 100
path = os.path.join(fn_temp_dir, 'input.json')
input_data = dict(manifest_file=manifest_file, resolution=resolution)
json_utilities.write(path, input_data)
cache = VoxelModelCache.from_json(path)
assert cache.manifest_file == manifest_file
assert cache.resolution == resolution
def main():
input_data = ju.read(INPUT_JSON)
structures = input_data.get('structures')
manifest_file = input_data.get('manifest_file')
manifest_file = os.path.join(TOP_DIR, manifest_file)
log_level = input_data.get('log_level', logging.DEBUG)
logging.getLogger().setLevel(log_level)
# experiments to exclude
experiments_exclude = ju.read(EXPERIMENTS_EXCLUDE_JSON)
eid_set = ju.read(EXPERIMENTS_PTP_JSON)
hyperparameters = ju.read(HYPERPARAMETER_JSON)
# get caching object
cache = VoxelModelCache(manifest_file=manifest_file)
output_dir = os.path.join(OUTPUT_DIR, 'voxel-%s' % ERROR_OPTION)
run_kwargs = dict(experiments_exclude=experiments_exclude,
error_option=ERROR_OPTION)
print(structures)
for structure in reversed(structures):
# get structure id
logging.debug("Running nested cross validation for structure: %s",
structure)
structure_id = get_structure_id(cache, structure)
run_kwargs.update(hyperparameters[structure])
scores = run_structure(cache, structure_id, eid_set=None, **run_kwargs)
logging.debug("voxel score : %.2f", scores['test_voxel'].mean())
logging.debug("regional score : %.2f", scores['test_regional'].mean())
write_output(output_dir, structure, structure_id, scores,
'scores_full')
logging.debug("Scoring only where power to predict")
try:
scores = run_structure(cache,
structure_id,
eid_set=eid_set,
**run_kwargs)
logging.debug("voxel score : %.2f", scores['test_voxel'].mean())
logging.debug("regional score : %.2f",
scores['test_regional'].mean())
except:
logging.debug("Not enough exps")
else:
write_output(output_dir, structure, structure_id, scores,
'scores_ptp')
def main():
input_data = ju.read(INPUT_JSON)
# experiments to exclude
experiments_exclude = ju.read(EXPERIMENTS_EXCLUDE_JSON)
manifest_file = input_data.get('manifest_file')
manifest_file = os.path.join(TOP_DIR, manifest_file)
log_level = input_data.get('log_level', logging.DEBUG)
logging.getLogger().setLevel(log_level)
# get cache, metric
logging.debug("loading experiments")
cache = VoxelModelCache(manifest_file=manifest_file)
weights = get_nnz_weights(cache, exp_exclude=experiments_exclude)
# get all weights
logging.debug("plotting")
plot_weights(weights)
def main():
input_data = ju.read(INPUT_JSON)
structures = input_data.get('structures')
manifest_file = input_data.get('manifest_file')
manifest_file = os.path.join(TOP_DIR, manifest_file)
log_level = input_data.get('log_level', logging.DEBUG)
logging.getLogger().setLevel(log_level)
# experiments to exclude
experiments_exclude = ju.read(EXPERIMENTS_EXCLUDE_JSON)
eid_set = ju.read(EXPERIMENTS_PTP_JSON)
# get caching object
cache = VoxelModelCache(manifest_file=manifest_file)
suffix = 'high_res' if HIGH_RES else 'standard'
output_dir = os.path.join(OUTPUT_DIR, 'homogeneous-%s' % suffix)
run_kwargs = dict(high_res=HIGH_RES, threshold_injection=THRESHOLD_INJECTION,
experiments_exclude=experiments_exclude, cv=CV)
for structure in structures:
# get structure id
logging.debug("Running nested cross validation for structure: %s", structure)
structure_id = get_structure_id(cache, structure)
scores = run_structure(cache, structure_id, eid_set=None, **run_kwargs)
logging.debug("regional score : %.2f", scores['test_regional'].mean())
write_output(output_dir, structure, structure_id, scores, 'scores_full')
logging.debug("Scoring only where power to predict")
try:
scores = run_structure(cache, structure_id, eid_set=eid_set, **run_kwargs)
logging.debug("regional score : %.2f", scores['test_regional'].mean())
except:
logging.debug("Not enough exps")
else:
write_output(output_dir, structure, structure_id, scores, 'scores_ptp')
def main(experiment_ids):
input_data = ju.read(INPUT_JSON)
manifest_file = input_data.get('manifest_file')
manifest_file = os.path.join(TOP_DIR, manifest_file)
log_level = input_data.get('log_level', logging.DEBUG)
logging.getLogger().setLevel(log_level)
# get voxel_model_cache
cache = VoxelModelCache(manifest_file=manifest_file)
# get model weights
logging.debug('loading model')
voxel_array, source_mask, target_mask = cache.get_voxel_connectivity_array()
# file save suffix
if SMOOTHED:
suffix = 'smoothed'
if LOG:
suffix += '-log'
elif LOG:
suffix = 'log'
else:
suffix = 'standard'
# top down viewer
tdv = TopDownView(cache, CMAP_FILE, blend_factor=BLEND_FACTOR)
if SMOOTHED:
osm = OptimizedSmoothedModel(cache, voxel_array, source_mask, target_mask)
for ss, experiment_ids in experiment_ids.items():
for eid in experiment_ids:
# get experiment projection/centroid
logging.debug('loading experiment %d', eid)
experiment = cache.get_projection_density(eid)[0]
logging.debug('getting model weights')
# ---------
# get model weights
centroid = tdv.get_experiment_centroid(eid)
if SMOOTHED:
logging.debug("Filtering target")
volume = osm.fit_voxel(centroid)
else:
row = source_mask.get_flattened_voxel_index(centroid)
volume = voxel_array[row]
if LOG:
logging.debug("inverse log transforming target")
volume = np.power(10.0, volume)
volume[volume < EPSILON] = 0 # instead of x-EPS for numerical reasons
model = target_mask.map_masked_to_annotation(volume)
# ---------
# get image
logging.debug('creating images')
exp = tdv.get_top_view(experiment)
mod = tdv.get_top_view(model)
# save
output_dir = os.path.join(OUTPUT_DIR, ss, str(eid))
if not os.path.exists(output_dir):
os.makedirs(output_dir)
logging.debug('saving')
exp.save(os.path.join(output_dir, 'data_%d.png' % eid))
mod.save(os.path.join(output_dir, '%s_model_%d.png' % (suffix, eid)))
out_data = dict(experiment_id=eid, centroid=centroid)
with open(os.path.join(output_dir, 'out_%d.json' % eid), 'w') as f:
json.dump(out_data, f, indent=2)
def main():
input_data = ju.read(INPUT_JSON)
structures = input_data.get('structures')
manifest_file = input_data.get('manifest_file')
manifest_file = os.path.join(TOP_DIR, manifest_file)
log_level = input_data.get('log_level', logging.DEBUG)
logging.getLogger().setLevel(log_level)
# experiments to exclude
experiments_exclude = ju.read(EXPERIMENTS_EXCLUDE_JSON)
# load hyperparameter dict
suffix = 'log' if LOG else 'standard'
hyperparameter_json = os.path.join(OUTPUT_DIR,
'hyperparameters-%s.json' % suffix)
hyperparameters = ju.read(hyperparameter_json)
# get caching object
cache = VoxelModelCache(manifest_file=manifest_file)
# get structure ids
structure_ids = [get_structure_id(cache, s) for s in structures]
# mask for reordering source
annotation = cache.get_annotation_volume()[0]
cumm_source_mask = np.zeros(annotation.shape, dtype=np.int)
offset = 1 # start @ 1 so that nonzero can be used
weights, nodes = [], []
for sid, sac in zip(structure_ids, structures):
logging.debug("Building model for structure: %s", sac)
data, reg = fit_structure(cache,
sid,
experiments_exclude,
hyperparameters[sac],
model_option=suffix)
w = reg.get_weights(data.injection_mask.coordinates)
# assign ordering to full source
ordering = np.arange(offset, w.shape[0] + offset, dtype=np.int)
offset += w.shape[0]
# get source mask
data.injection_mask.fill_volume_where_masked(cumm_source_mask,
ordering)
# append to list
weights.append(w)
nodes.append(reg.nodes)
# stack
weights = padded_diagonal_fill(weights)
nodes = np.vstack(nodes)
# need to reorder weights
# (subtract 1 to get proper index)
permutation = cumm_source_mask[cumm_source_mask.nonzero()] - 1
weights = weights[permutation, :]
# regionalized
logging.debug('regionalizing voxel weights')
ontological_order = get_ordered_summary_structures(cache)
source_mask = Mask.from_cache(cache,
structure_ids=structure_ids,
hemisphere_id=2)
source_key = source_mask.get_key(structure_ids=ontological_order)
ipsi_key = data.projection_mask.get_key(structure_ids=ontological_order,
hemisphere_id=2)
contra_key = data.projection_mask.get_key(structure_ids=ontological_order,
hemisphere_id=1)
ipsi_model = RegionalizedModel(weights,
nodes,
source_key,
ipsi_key,
ordering=ontological_order,
dataframe=True)
contra_model = RegionalizedModel(weights,
nodes,
source_key,
contra_key,
ordering=ontological_order,
dataframe=True)
get_metric = lambda s: pd.concat(
(getattr(ipsi_model, s), getattr(contra_model, s)),
keys=('ipsi', 'contra'),
axis=1)
# write results
output_dir = os.path.join(TOP_DIR, 'connectivity',
'voxel-%s-model' % suffix)
if not os.path.exists(output_dir):
os.makedirs(output_dir)
# regionalized
logging.debug('saving to directory: %s', output_dir)
get_metric('connection_density').to_csv(
os.path.join(output_dir, 'connection_density.csv'))
get_metric('connection_strength').to_csv(
os.path.join(output_dir, 'connection_strength.csv'))
get_metric('normalized_connection_density').to_csv(
os.path.join(output_dir, 'normalized_connection_density.csv'))
get_metric('normalized_connection_strength').to_csv(
os.path.join(output_dir, 'normalized_connection_strength.csv'))
# voxel
ju.write(os.path.join(output_dir, 'target_mask_params.json'),
dict(structure_ids=structure_ids, hemisphere_id=3))
ju.write(os.path.join(output_dir, 'source_mask_params.json'),
dict(structure_ids=structure_ids, hemisphere_id=2))
np.savetxt(os.path.join(output_dir, 'weights.csv.gz'),
weights.astype(np.float32),
delimiter=',')
np.savetxt(os.path.join(output_dir, 'nodes.csv.gz'),
nodes.astype(np.float32),
delimiter=',')
def main():
# caching object for downloading/loading connectivity/model data
cache = VoxelModelCache(manifest_file=MANIFEST_FILE)
# load voxel model
logging.info('loading voxel array')
voxel_array, source_mask, target_mask = cache.get_voxel_connectivity_array(
)
reference_shape = source_mask.reference_space.annotation.shape
vmax = 1.2 * np.percentile(voxel_array.nodes, 99)
# 2D Cortical Surface Mapper
mapper = CorticalMap(projection='top_view')
# colormaps
cmap_view = matplotlib.cm.viridis
cmap_pixel = matplotlib.cm.spring
cmap_view.set_bad(alpha=0)
cmap_pixel.set_bad(alpha=0)
# only want R hemisphere
lookup = mapper.view_lookup.copy().T # transpose for vertical pixel query
lookup[:lookup.shape[0] // 2, :] = -1
# dict(2D lookup_value -> avg(path))
logging.info('beginning image creation')
for i, val in enumerate(lookup[lookup > -1]):
# get the mean path voxel
path = mapper.paths[val][mapper.paths[val].nonzero()]
path = np.vstack([np.unravel_index(x, reference_shape) for x in path])
voxel = tuple(map(int, path.mean(axis=0)))
try:
row_idx = source_mask.get_flattened_voxel_index(voxel)
except ValueError:
logging.warning('voxel %s not in mask', voxel)
else:
# get voxel expression
volume = target_mask.fill_volume_where_masked(
np.zeros(reference_shape), voxel_array[row_idx])
# map to cortical surface
flat_view = mapper.transform(volume, fill_value=np.nan)
plt.pcolormesh(flat_view,
zorder=1,
cmap=cmap_view,
vmin=0,
vmax=vmax)
# injection location
pixel = np.ma.masked_where(mapper.view_lookup != val,
flat_view,
copy=False)
plt.pcolormesh(pixel, zorder=2, cmap=cmap_pixel, vmin=0, vmax=1)
# plot params
plt.gca().invert_yaxis()
plt.axis('off')
plt.savefig(os.path.join(OUTPUT_DIR, '%05d.png' % i),
bbox_inches='tight',
facecolor=None,
edgecolor=None,
transparent=True,
dpi=50)
plt.close()
logging.info('converting images to gif')
subprocess.run(GIF_CONVERT_COMMAND,
stdout=subprocess.PIPE,
cwd=OUTPUT_DIR,
shell=True)
def main():
# caching object for downloading/loading connectivity/model data
cache = VoxelModelCache(manifest_file=MANIFEST_FILE)
# load voxel model
logging.info('loading voxel array')
voxel_array, source_mask, target_mask = cache.get_voxel_connectivity_array(
)
reference_shape = source_mask.reference_space.annotation.shape
vmax = 1.2 * np.percentile(voxel_array.nodes, 99)
# 2D Cortical Surface Mapper
# projection: can change to "flatmap" if desired
mapper = CorticalMap(projection='top_view')
# quick hack to fix bug
mapper.view_lookup[51, 69] = mapper.view_lookup[51, 68]
mapper.view_lookup[51, 44] = mapper.view_lookup[51, 43]
# colormaps
cmap_view = matplotlib.cm.inferno
cmap_pixel = matplotlib.cm.cool
cmap_view.set_bad(alpha=0)
cmap_pixel.set_bad(alpha=0)
# only want R hemisphere
lookup = mapper.view_lookup.copy().T # transpose for vertical pixel query
lookup[:lookup.shape[0] // 2, :] = -1
# dict(2D lookup_value -> avg(path))
logging.info('beginning image creation')
for i, val in enumerate(lookup[lookup > -1]):
# get the mean path voxel
print("Evaluating pixel %d" % i)
path = mapper.paths[val][mapper.paths[val].nonzero()]
path = np.vstack([np.unravel_index(x, reference_shape) for x in path])
voxel = tuple(map(int, path.mean(axis=0)))
try:
row_idx = source_mask.get_flattened_voxel_index(voxel)
except ValueError:
logging.warning('voxel %s not in mask', voxel)
else:
# get voxel expression
volume = target_mask.fill_volume_where_masked(
np.zeros(reference_shape), voxel_array[row_idx])
# map to cortical surface
flat_view = mapper.transform(volume, fill_value=np.nan)
# injection location
pixel = np.ma.masked_where(mapper.view_lookup != val,
flat_view,
copy=False)
# plot & params
fig, ax = plt.subplots(figsize=(6, 6))
# plot connectivity
im = plt.pcolormesh(flat_view,
zorder=1,
cmap=cmap_view,
vmin=0,
vmax=vmax)
# plot source voxel
plt.pcolormesh(pixel, zorder=2, cmap=cmap_pixel, vmin=0, vmax=1)
plt.gca().invert_yaxis() # flips yaxis
plt.axis('off')
# plot overlay
extent = plt.gca().get_xlim() + plt.gca().get_ylim()
plt.imshow(top_down_overlay,
interpolation="nearest",
extent=extent,
zorder=3)
# add colorbar
cbar = plt.colorbar(im,
shrink=0.3,
use_gridspec=True,
format="%1.1e")
cbar.set_ticks([0, 0.0004, 0.0008, 0.0012])
cbar.ax.tick_params(labelsize=6)
plt.tight_layout()
plt.savefig(os.path.join(OUTPUT_DIR, '%05d.png' % i),
bbox_inches=None,
facecolor=None,
edgecolor=None,
transparent=True,
dpi=240)
plt.close()
class VolumetricAPI(Paths):
"""
This class takes care of downloading, analysing and rendering data from:
"High-resolution data-driven model of the mouse connectome ", Knox et al 2018.
[https://www.mitpressjournals.org/doi/full/10.1162/netn_a_00066].
These data can be used to look at spatialised projection strength with sub-region (100um) resolution.
e.g. to look at where in region B are the projections from region A, you can use this class.
To download the data, this class uses code from: https://github.com/AllenInstitute/mouse_connectivity_models.
"""
voxel_size = 100
projections = {}
mapped_projections = {}
hemispheres = dict(left=1, right=2, both=3)
def __init__(self,
base_dir=None,
add_root=True,
use_cache=True,
scene_kwargs={},
**kwargs):
"""
Initialise the class instance to get a few useful paths and variables.
:param base_dir: str, path to base directory in which all of brainrender data are stored.
Pass only if you want to use a different one from what's default.
:param add_root: bool, if True the root mesh is added to the rendered scene
:param use_cache: if true data are loaded from a cache to speed things up.
Useful to set it to false to help debugging.
:param scene_kwargs: dict, params passed to the instance of Scene associated with this class
"""
Paths.__init__(self, base_dir=base_dir, **kwargs)
# Get MCM cache
cache_path = os.path.join(self.mouse_connectivity_volumetric,
'voxel_model_manifest.json')
if not os.path.isfile(cache_path):
if not connected_to_internet():
raise ValueError(
"The first time you use this class it will need to download some data, but it seems that you're not connected to the internet."
)
print(
"Downloading volumetric data. This will take several minutes but it only needs to be done once."
)
self.cache = VoxelModelCache(manifest_file=cache_path)
self.voxel_array = None
self.target_coords, self.source_coords = None, None
# Get projection cache paths
self.data_cache = self.mouse_connectivity_volumetric_cache
self.data_cache_projections = os.path.join(self.data_cache,
"projections")
self.data_cache_targets = os.path.join(self.data_cache, "targets")
self.data_cache_sources = os.path.join(self.data_cache, "sources")
for fold in [
self.data_cache_projections, self.data_cache_targets,
self.data_cache_sources
]:
if not os.path.isdir(fold):
os.mkdir(fold)
# Get structures tree
self.structure_tree = self.cache.get_structure_tree()
# Get scene
self.scene = Scene(add_root=add_root, **scene_kwargs)
# Other vars
self.use_cache = use_cache
def __getattr__(self, attr):
__dict__ = super(VolumetricAPI, self).__getattribute__('__dict__')
try:
return __dict__['scene'].__getattribute__(attr)
except AttributeError as e:
raise AttributeError(
f"Could not attribute {attr} for class VolumetricAPI:\n{e}")
# ---------------------------------------------------------------------------- #
# UTILS #
# ---------------------------------------------------------------------------- #
# ------------------------- Interaction with mcmodels ------------------------ #
def _get_structure_id(self, struct):
" Get the ID of a structure (or list of structures) given it's acronym"
if not isinstance(struct, (list, tuple)):
struct = [struct]
return [
self.structure_tree.get_structures_by_acronym([s])[0]["id"]
for s in struct
]
def _load_voxel_data(self):
"Load the VoxelData array from Knox et al 2018"
if self.voxel_array is None:
# Get VoxelArray
weights_file = os.path.join(self.mouse_connectivity_volumetric,
'voxel_model', 'weights')
nodes_file = os.path.join(self.mouse_connectivity_volumetric,
'voxel_model', 'nodes')
# Try to load from numpy
if os.path.isfile(weights_file + '.npy.gz'):
weights = load_npy_from_gz(weights_file + '.npy.gz')
nodes = load_npy_from_gz(nodes_file + '.npy.gz')
# Create array
self.voxel_array = VoxelConnectivityArray(weights, nodes)
# Get target and source masks
self.source_mask = self.cache.get_source_mask()
self.target_mask = self.cache.get_target_mask()
else:
print("Loading voxel data, might take a few minutes.")
# load from standard cache
self.voxel_array, self.source_mask, self.target_mask = self.cache.get_voxel_connectivity_array(
)
# save to npy
save_npy_to_gz(weights_file + '.npy.gz',
self.voxel_array.weights)
save_npy_to_gz(nodes_file + '.npy.gz', self.voxel_array.nodes)
def _get_coordinates_from_voxel_id(self, p0, as_source=True):
"""
Takes the index of a voxel and returns the 3D coordinates in reference space.
The index number should be extracted with either a source_mask or a target_mask.
If target_mask wa used set as_source as False.
:param p0: int
"""
if self.voxel_array is None:
self._load_voxel_data()
if as_source:
return self.source_mask.coordinates[p0] * self.voxel_size
else:
return self.target_mask.coordinates[p0] * self.voxel_size
def _get_mask_coords(self, as_source):
if as_source:
if self.source_coords is None:
coordinates = self.source_mask.coordinates * self.voxel_size
self.source_coords = coordinates
else:
coordinates = self.source_coords
else:
if self.target_coords is None:
coordinates = self.target_mask.coordinates * self.voxel_size
self.target_coords = coordinates
else:
coordinates = self.target_coords
return coordinates
def _get_voxel_id_from_coordinates(self, p0, as_source=True):
if self.voxel_array is None:
self._load_voxel_data()
# Get the brain region from the coordinates
coordinates = self._get_mask_coords(as_source)
# Get the position of p0 in the coordinates volumetric array
p0 = np.int64([round(p, -2) for p in p0])
try:
x_idx = (np.abs(coordinates[:, 0] - p0[0])).argmin()
y_idx = (np.abs(coordinates[:, 1] - p0[1])).argmin()
z_idx = (np.abs(coordinates[:, 2] - p0[2])).argmin()
p0_idx = [x_idx, y_idx, z_idx]
except:
raise ValueError(
f"Could not find the voxe corresponding to the point given: {p0}"
)
return p0_idx[0]
# ----------------------------------- Cache ---------------------------------- #
def _get_cache_filename(self, tgt, what):
"""Data are cached according to a naming convention, this function gets the name for an object
according to the convention"""
if what == 'projection':
fld = self.data_cache_projections
elif what == 'source':
fld = self.data_cache_sources
elif what == 'target':
fld = self.data_cache_targets
else:
raise ValueError(
f'Error while getting cached data file name.\n' +
f'What was {what} but should be projection/source/target/actor.'
)
name = ''.join([str(i) for i in tgt])
path = os.path.join(fld, name + '.npy.gz')
return name, path, os.path.isfile(path)
def _get_from_cache(self, tgt, what):
""" tries to load objects from cached data, if they exist"""
if not self.use_cache:
return None
name, cache_path, cache_exists = self._get_cache_filename(tgt, what)
if not cache_exists:
return None
else:
return load_npy_from_gz(cache_path)
def save_to_cache(self, tgt, what, obj):
""" Saves data to cache to avoid loading thema again in the future"""
name, cache_path, _ = self._get_cache_filename(tgt, what)
save_npy_to_gz(cache_path, obj)
# ---------------------------------------------------------------------------- #
# PREPROCESSING #
# ---------------------------------------------------------------------------- #
# ------------------------- Sources and targets masks ------------------------ #
def get_source(self, source, hemisphere='both'):
"""
Loads the mask for a source structure
:param source: str or list of str with acronym of source regions
:param hemisphere: str, ['both', 'left', 'right']. Which hemisphere to consider.
"""
if not isinstance(source, (list, tuple)):
source = [source]
self.source = self._get_from_cache(source, 'source')
if self.source is None:
self._load_voxel_data()
source_ids = self._get_structure_id(source)
self.source = self.source_mask.get_structure_indices(
structure_ids=source_ids,
hemisphere_id=self.hemispheres[hemisphere])
self.save_to_cache(source, 'source', self.source)
return self.source
def get_target_mask(self, target, hemisphere):
"""returns a 'key' array and a mask object
used to transform projection data from linear arrays to 3D volumes.
"""
target_ids = self._get_structure_id(target)
self.tgt_mask = Mask.from_cache(
self.cache,
structure_ids=target_ids,
hemisphere_id=self.hemispheres[hemisphere])
def get_target(self, target, hemisphere='both'):
"""
Loads the mask for a target structure.
:param target: str or list of str with acronym of target regions
:param hemisphere: str, ['both', 'left', 'right']. Which hemisphere to consider.
"""
if not isinstance(target, (list, tuple)):
target = [target]
if hemisphere != 'both':
cache_name = target + [hemisphere]
else:
cache_name = target
self.target = self._get_from_cache(cache_name, 'target')
if self.target is None:
self._load_voxel_data()
target_ids = self._get_structure_id(target)
self.target = self.target_mask.get_structure_indices(
structure_ids=target_ids,
hemisphere_id=self.hemispheres[hemisphere])
self.save_to_cache(cache_name, 'target', self.target)
return self.target
# -------------------------------- Projections ------------------------------- #
def get_projection(self,
source,
target,
name,
hemisphere='both',
projection_mode='mean',
mode='target'):
"""
Gets the spatialised projection intensity from a source to a target.
:param source: str or list of str with acronym of source regions
:param target: str or list of str with acronym of target regions
:param name: str, name of the projection
:param projection_mode: str, if 'mean' the data from different experiments are averaged,
if 'max' the highest value is taken.
:param mode: str. If 'target' the spatialised projection strength in the target structures is returned, usefule
to see where source projects to in target. Otherwise if 'source' the spatialised projection strength in
the source structure is return. Useful to see which part of source projects to target.
:return: 1D numpy array with mean projection from source to target voxels
"""
if mode == 'target':
self.get_target_mask(target, hemisphere)
elif mode == 'source':
self.get_target_mask(source, 'right')
else:
raise ValueError(
f'Invalide mode: {mode}. Should be either source or target.')
cache_name = sorted(source) + ['_'] + sorted(target) + [
f'_{projection_mode}_{mode}'
]
if hemisphere != 'both':
cache_name += [hemisphere]
proj = self._get_from_cache(cache_name, 'projection')
if proj is None:
source_idx = self.get_source(source, hemisphere)
target_idx = self.get_target(target, hemisphere)
self._load_voxel_data()
projection = self.voxel_array[source_idx, target_idx]
if mode == 'target':
axis = 0
elif mode == 'source':
axis = 1
else:
raise ValueError(
f'Invalide mode: {mode}. Should be either source or target.'
)
if projection_mode == 'mean':
proj = np.mean(projection, axis=axis)
elif projection_mode == 'max':
proj = np.max(projection, axis=axis)
else:
raise ValueError(
f'Projection mode {projection_mode} not recognized.\n' +
'Should be one of: ["mean", "max"].')
# Save to cache
self.save_to_cache(cache_name, 'projection', proj)
self.projections[name] = proj
return proj
def get_mapped_projection(self, source, target, name, **kwargs):
"""
Gets the spatialised projection intensity from a source to a target, but as
a mapped volume instead of a linear array.
:param source: str or list of str with acronym of source regions
:param target: str or list of str with acronym of target regions
:param name: str, name of the projection
:return: 3D numpy array with projectino intensity
"""
projection = self.get_projection(source, target, name, **kwargs)
mapped_projection = self.tgt_mask.map_masked_to_annotation(projection)
self.mapped_projections[name] = mapped_projection
return mapped_projection
def get_mapped_projection_to_point(self,
p0,
restrict_to=None,
restrict_to_hemisphere='both'):
"""
Gets projection intensity from all voxels to the voxel corresponding to a point of interest
"""
cache_name = f'proj_to_{p0[0]}_{p0[1]}_{p0[1]}'
if restrict_to is not None:
cache_name += f'_{restrict_to}'
proj = self._get_from_cache(cache_name, 'projection')
if proj is None:
p0idx = self._get_voxel_id_from_coordinates(p0, as_source=False)
if restrict_to is not None:
source_idx = self.get_source(restrict_to,
restrict_to_hemisphere)
proj = self.voxel_array[source_idx, p0idx]
self.get_target_mask(restrict_to, restrict_to_hemisphere)
mapped_projection = self.tgt_mask.map_masked_to_annotation(
proj)
else:
proj = self.voxel_array[:, p0idx]
mapped_projection = self.source_mask.map_masked_to_annotation(
proj)
self.save_to_cache(cache_name, 'projection', mapped_projection)
return mapped_projection
else:
return proj
def get_mapped_projection_from_point(self,
p0,
restrict_to=None,
restrict_to_hemisphere='both'):
"""
Gets projection intensity from all voxels to the voxel corresponding to a point of interest
"""
if self.get_hemispere_from_point(p0) == 'left':
raise ValueError(
f'The point passed [{p0}] is in the left hemisphere,' +
' but "projection from point" only works from the right hemisphere.'
)
cache_name = f'proj_from_{p0[0]}_{p0[1]}_{p0[1]}'
if restrict_to is not None:
cache_name += f'_{restrict_to}'
proj = self._get_from_cache(cache_name, 'projection')
if proj is None:
p0idx = self._get_voxel_id_from_coordinates(p0, as_source=True)
if restrict_to is not None:
target_idx = self.get_target(restrict_to,
restrict_to_hemisphere)
proj = self.voxel_array[p0idx, target_idx]
self.get_target_mask(restrict_to, restrict_to_hemisphere)
mapped_projection = self.tgt_mask.map_masked_to_annotation(
proj)
else:
proj = self.voxel_array[p0idx, :]
mapped_projection = self.target_mask.map_masked_to_annotation(
proj)
self.save_to_cache(cache_name, 'projection', mapped_projection)
return mapped_projection
else:
return proj
# ---------------------------------------------------------------------------- #
# RENDERING #
# ---------------------------------------------------------------------------- #
def add_mapped_projection(self,
source,
target,
actor_kwargs={},
render_source_region=False,
render_target_region=False,
regions_kwargs={},
**kwargs):
"""
Gets the spatialised projection intensity from a source to a target
and renders it as a vtkplotter lego visualisation.
:param source: str or list of str with acronym of source regions
:param target: str or list of str with acronym of target regions
:param render_source_region: bool, if true a wireframe mesh of source regions is rendered
:param render_target_region: bool, if true a wireframe mesh of target regions is rendered
:param regions_kwargs: pass options to specify how brain regions should look like
:param kwargs: kwargs can be used to control how the rendered object looks like.
Look at the arguments of 'add_volume' to see what arguments are available.
"""
# Get projection data
if not isinstance(source, list): source = [source]
if not isinstance(target, list): target = [target]
name = ''.join(source) + '_'.join(target)
mapped_projection = self.get_mapped_projection(source, target, name,
**kwargs)
lego_actor = self.add_volume(mapped_projection, **actor_kwargs)
# Render relevant regions meshes
if render_source_region or render_target_region:
wireframe = regions_kwargs.pop('wireframe', True)
use_original_color = regions_kwargs.pop('use_original_color', True)
if render_source_region:
self.scene.add_brain_regions(
source,
use_original_color=use_original_color,
wireframe=wireframe,
**regions_kwargs)
if render_target_region:
self.scene.add_brain_regions(
target,
use_original_color=use_original_color,
wireframe=wireframe,
**regions_kwargs)
return lego_actor
def add_mapped_projection_to_point(self,
p0,
show_point=True,
show_point_region=False,
show_crosshair=True,
crosshair_kwargs={},
point_region_kwargs={},
point_kwargs={},
from_point=False,
**kwargs):
if not isinstance(p0, (list, tuple, np.ndarray)):
raise ValueError(
"point passed should be a list or a 1d array, not: {p0}")
restrict_to = kwargs.pop('restrict_to', None)
restrict_to_hemisphere = kwargs.pop('restrict_to_hemisphere', 'both')
if not from_point:
projection = self.get_mapped_projection_to_point(
p0,
restrict_to=restrict_to,
restrict_to_hemisphere=restrict_to_hemisphere)
else:
projection = self.get_mapped_projection_from_point(
p0,
restrict_to=restrict_to,
restrict_to_hemisphere=restrict_to_hemisphere)
lego_actor = self.add_volume(projection, **kwargs)
if show_point:
color = point_kwargs.pop('color', 'salmon')
radius = point_kwargs.pop('radius', 50)
alpha = point_kwargs.pop('alpha', 1)
if not show_crosshair:
self.scene.add_sphere_at_point(p0,
color=color,
radius=radius,
alpha=alpha,
**point_kwargs)
else:
ml = crosshair_kwargs.pop('ml', True)
dv = crosshair_kwargs.pop('dv', True)
ap = crosshair_kwargs.pop('ap', True)
self.scene.add_crosshair_at_point(p0,
ml=ml,
dv=dv,
ap=ap,
line_kwargs=crosshair_kwargs,
point_kwargs={
'color': color,
'radius': radius,
'alpha': alpha
})
if show_point_region:
use_original_color = point_region_kwargs.pop(
'use_original_color', False)
alpha = point_region_kwargs.pop('alpha', 0.3)
region = self.scene.get_structure_from_coordinates(p0)
self.scene.add_brain_regions([region],
use_original_color=use_original_color,
alpha=alpha,
**point_region_kwargs)
return lego_actor
def add_mapped_projection_from_point(self, *args, **kwargs):
return self.add_mapped_projection_to_point(*args,
**kwargs,
from_point=True)
def add_volume(self,
volume,
cmap='afmhot_r',
alpha=1,
add_colorbar=True,
**kwargs):
"""
Renders intensitdata from a 3D numpy array as a lego volumetric actor.
:param volume: np 3D array with number of dimensions = those of the 100um reference space.
:param cmap: str with name of colormap to use
:param alpha: float, transparency
:param add_colorbar: if True a colorbar is added to show the values of the colormap
"""
# Parse kwargs
line_width = kwargs.pop('line_width', 1)
if cmap == 'random' or not cmap or cmap is None:
cmap = get_random_colormap()
# Get vmin and vmax threshold for visualisation
vmin = kwargs.pop('vmin', 0.000001)
vmax = kwargs.pop('vmax', np.nanmax(volume))
# Check values
if np.max(volume) > vmax:
print(
"While rendering mapped projection some of the values are above the vmax threshold."
+ "They will not be displayed." +
f" vmax was {vmax} but found value {round(np.max(volume), 5)}."
)
if vmin > vmax:
raise ValueError(
f'The vmin threhsold [{vmin}] cannot be larger than the vmax threshold [{vmax}'
)
if vmin < 0: vmin = 0
# Get 'lego' actor
vol = Volume(volume)
lego = vol.legosurface(vmin=vmin, vmax=vmax, cmap=cmap)
# Scale and color actor
lego.alpha(alpha).lw(line_width).scale(self.voxel_size)
lego.cmap = cmap
# Add colorbar
if add_colorbar:
lego.addScalarBar(vmin=vmin,
vmax=vmax,
horizontal=1,
c='k',
pos=(0.05, 0.05),
titleFontSize=40)
# Add to scene
actor = self.scene.add_vtkactor(lego)
return actor
def main():
# initialize cache object
cache = VoxelModelCache(manifest_file=MANIFEST_FILE)
rs = cache.get_reference_space()
rs.remove_unassigned(update_self=True)
# major id children only
structures = []
for s in rs.structure_tree.get_structures_by_set_id([MAJOR_BRAIN_SET_ID]):
structures.extend(rs.structure_tree.descendants([s['id']])[0])
# load in voxel model
print('loading array')
voxel_array, source_mask, target_mask = cache.get_voxel_connectivity_array(
)
print('getting keys')
source_keys, target_keys = [], []
source_counts, target_counts = [], []
for s in structures:
s_mask = source_mask.reference_space.make_structure_mask(
[s['id']], direct_only=False)
t_mask = target_mask.reference_space.make_structure_mask(
[s['id']], direct_only=False)
# NOTE: ipsi
t_mask[..., :t_mask.shape[-1] // 2] = 0
# keys
s_key = source_mask.mask_volume(s_mask).nonzero()[0]
t_key = target_mask.mask_volume(t_mask).nonzero()[0]
source_keys.append(s_key)
target_keys.append(t_key)
source_counts.append(s_key.size)
target_counts.append(t_key.size)
del source_mask
del target_mask
# arrays
source_counts = np.asarray(source_counts)
target_counts = np.asarray(target_counts)
# compute
print('computing regional')
connection_strength = np.empty(2 * [len(structures)])
for i, j in itertools.product(range(len(structures)), repeat=2):
print(i, j)
connection_strength[i, j] = voxel_array[source_keys[i],
target_keys[j]].sum()
del voxel_array
del source_keys
del target_keys
structure_acronyms = [s['acronym'] for s in structures]
connection_strength = pd.DataFrame(connection_strength,
index=structure_acronyms,
columns=structure_acronyms)
# other metrics
connection_density = np.divide(connection_strength,
source_counts[:, np.newaxis])
normalized_connection_strength = np.divide(connection_strength,
target_counts[:, np.newaxis])
normalized_connection_density = np.divide(
connection_strength, np.outer(source_counts, target_counts))
# save
connection_strength.to_csv(
os.path.join(OUTPUT_DIR, 'connection_strength.csv'))
connection_density.to_csv(
os.path.join(OUTPUT_DIR, 'connection_density.csv'))
normalized_connection_strength.to_csv(
os.path.join(OUTPUT_DIR, 'normalized_connection_strength.csv'))
normalized_connection_density.to_csv(
os.path.join(OUTPUT_DIR, 'normalized_connection_density.csv'))
if __name__ == "__main__":
input_data = ju.read(INPUT_JSON)
manifest_file = input_data.get('manifest_file')
manifest_file = os.path.join(TOP_DIR, manifest_file)
log_level = input_data.get('log_level', logging.DEBUG)
logging.getLogger().setLevel(log_level)
# configure
colors = sns.color_palette(n_colors=2)
# get cache, metric
cache = VoxelModelCache(manifest_file=manifest_file)
df_metric = cache.get_normalized_connection_density(dataframe=True)
logging.debug("getting cortical network")
df_cortex = get_cortical_df(df_metric, cache)
# region acs
region_acs = df_metric.index.values
logging.debug("computing distances")
d = get_distances(region_acs, cache)
d = to_dataframe(d, df_metric.index, df_metric.columns)
d_cortex = get_cortical_df(d, cache)
# get projection types
full_ipsi = get_pt((d, df_metric), thresh=0)
