def _generate_subject(sid, infmaps, bdists):
import neuropythy as ny, six
lh = VisualPerformanceFieldsDataset._generate_hemi(
sid, 'lh', infmaps, bdists)
rh = VisualPerformanceFieldsDataset._generate_hemi(
sid, 'rh', infmaps, bdists)
return ny.hcp_subject(sid).with_hemi(lh=lh, rh=rh)
def _generate_hemi(sid, h, infmaps, bdists):
import neuropythy as ny, six
hem = ny.hcp_subject(sid).hemis[h]
bi = infmaps[sid][h]
ds = bdists[sid][h]
ps = {}
for (k, v) in six.iteritems(ds):
ps[k + '_distance'] = v
for (k, v) in six.iteritems(bi):
ps[k] = v
return hem.with_prop(ps)
def _generate_boundary_distances(sid, h, infmaps):
import neuropythy as ny
import numpy as np
import six
sub = neuropythy.hcp_subject(sid)
hemi = sub.hemis[h]
# we use the inferred retinotopy data loaded above
rdat = ny.retinotopy_data(infmaps[sid][h], 'inf_')
# get x/y and labels
(x0, y0) = ny.as_retinotopy(rdat, 'geographical')
lbl = rdat['visual_area']
(ang, ecc) = ny.as_retinotopy(rdat, 'visual')
# get V1/V2 lines:
mask = (ecc <= 7) & (lbl > 0)
v1 = np.where(mask & (lbl == 1))[0]
v2 = np.where(mask & (lbl == 2))[0]
v12 = np.union1d(v1, v2)
# invert x for v2
x = np.array(x0)
y = y0
x[v2] = -x[v2]
# make a visual field mesh
vmesh = hemi.surfaces['midgray'].submesh(v12)
vmesh = vmesh.copy(coordinates=(x[vmesh.labels], y[vmesh.labels]))
# draw a line from origin to the end of the stimulus; see what intersects it
addrs = {}
paths = {}
for (ept, nm) in zip([(0, 7), (0, -7), (7, 0)],
['ventral', 'dorsal', 'horizontal']):
seg = np.array([(0, 0), ept])
ipts = ny.geometry.segment_intersection_2D(seg,
vmesh.edge_coordinates)
ipts = np.transpose(ipts)
ipts = ipts[np.isfinite(ipts[:, 0])]
# these are the points to address
dists = np.dot(ipts, seg[1] - seg[0])
ipts = ipts[np.argsort(dists)]
# possibly the first/last few are not in the mesh; just ignore these
addr = vmesh.address(ipts)
ii = np.isfinite(addr['coordinates'][0])
addr = {k: v[:, ii] for (k, v) in six.iteritems(addr)}
addrs[nm] = addr
paths[nm] = ny.geometry.Path(hemi, addr)
# use these paths to calculate distances
eds = {
k: p.estimated_distances['midgray']
for (k, p) in six.iteritems(paths)
}
return eds
def _generate_summary_table(infmaps, bdists, subject_list):
import neuropythy as ny
import six, numpy as np
cols = [
'sid', 'hemi', 'prf_polar_angle', 'prf_eccentricity',
'prf_variance_explained', 'prf_radius', 'visual_area',
'inf_polar_angle', 'inf_eccentricity', 'ventral', 'dorsal',
'horizontal_distance', 'ventral_distance', 'dorsal_distance'
]
df = {k: [] for k in cols}
for sid in subject_list:
sub = ny.hcp_subject(sid)
for h in ('lh', 'rh'):
hemi = sub.hemis[h]
dists = bdists[sid][h]
if dists is None or any(v is None
for v in six.itervalues(dists)):
continue
rmaps = ny.retinotopy_data(hemi, 'prf_')
infmap = infmaps[sid][h]
lbls = np.array(infmap['inf_visual_area'])
iang = np.array(infmap['inf_polar_angle'])
iecc = np.array(infmap['inf_eccentricity'])
mask = np.where(np.isin(lbls, [1, 2]) & (iecc <= 7))[0]
vnt = (np.abs(iang) < 90)
drs = ~vnt
n = len(mask)
dat = {
'sid': np.full(n, sid),
'hemi': np.full(n, h),
'prf_polar_angle': rmaps['polar_angle'][mask],
'prf_eccentricity': rmaps['eccentricity'][mask],
'prf_radius': rmaps['radius'][mask],
'prf_variance_explained':
rmaps['variance_explained'][mask],
'visual_area': lbls[mask],
'inf_polar_angle': np.abs(iang[mask]),
'inf_eccentricity': iecc[mask],
'ventral': vnt[mask],
'dorsal': drs[mask],
'horizontal_distance': dists['horizontal'][mask],
'ventral_distance': dists['ventral'][mask],
'dorsal_distance': dists['dorsal'][mask]
}
for (k, v) in six.iteritems(dat):
df[k].append(v)
df = {k: np.concatenate(v) for (k, v) in six.iteritems(df)}
return ny.to_dataframe(df)
def load_hcp_data(credentials, slice_number):
# Import dictionaries
import neuropythy as ny
import ipyvolume as ipv
import nibabel as nib
import numpy as np
# Configure neuropythy
ny.config['hcp_credentials'] = credentials
fs = ny.data['hcp'].s3fs
# Get full path to T1w NIFTIs
fid = fs.ls('hcp-openaccess/HCP_1200/')
fid.pop(0)
fpath = []
for f in fid:
fpath.append(f + '/MNINonLinear/T1w.nii.gz')
# Get list of subject IDs
sid = []
for f in fid:
sid.append(f.split('/')[2])
# Save single slice for each subject into concatenated array
im_array = []
sid_tmp = sid[0:3]
for i in sid_tmp:
print(i)
sub = ny.hcp_subject(i)
im = sub.load('MNINonLinear/T1w.nii.gz')
data = im.get_fdata()
im_array.append(data[:, :, slice_number])
arr = np.asarray(im_array)
arr_padded = np.pad(img, [(0, 0), (25, 26), (0, 0)],
mode='constant') # hard coded to pad this dimension!
return arr, sid