def load_surf_bold_mask(bold_files, mesh_file, mask_file=None):
from pyhrf.tools._io import read_mesh, read_texture, discard_bad_data
logger.info('Load mesh: ' + mesh_file)
coords, triangles, coord_sys = read_mesh(mesh_file)
logger.debug('Build graph...')
fgraph = graph_from_mesh(triangles)
assert graph_is_sane(fgraph)
logger.info('Mesh has %d nodes', len(fgraph))
logger.debug('Compute length of edges ... ')
edges_l = np.array([np.array([distance(coords[i],
coords[n],
coord_sys.xform)
for n in nl])
for i, nl in enumerate(fgraph)], dtype=object)
if mask_file is None or not op.exists(mask_file):
logger.warning('Mask file %s does not exist. Taking '
'all nodes ...', mask_file)
mask = np.ones(len(fgraph))
mask_meta_obj = None
mask_loaded_from_file = False
else:
mask, mask_meta_obj = read_texture(mask_file)
mask_loaded_from_file = True
if not (np.round(mask) == mask).all():
raise Exception("Mask is not n-ary")
if len(mask) != len(fgraph):
raise Exception('Size of mask (%d) is different from size '
'of graph (%d)' % (len(mask), len(fgraph)))
mask = mask.astype(np.int32)
if mask.min() == -1:
mask += 1
# Split graph into rois:
graphs = {}
edge_lengths = {}
for roi_id in np.unique(mask):
mroi = np.where(mask == roi_id)
graph, _ = sub_graph(fgraph, mroi[0])
edge_lengths[roi_id] = edges_l[mroi]
graphs[roi_id] = graph
# Load BOLD:
last_scan = 0
session_scans = []
bolds = []
for bold_file in bold_files:
logger.info('load bold: %s', bold_file)
bold, _ = read_texture(bold_file)
logger.info('bold shape: %s', str(bold.shape))
bolds.append(bold)
session_scans.append(np.arange(last_scan, last_scan + bold.shape[0],
dtype=int))
last_scan += bold.shape[0]
bold = np.concatenate(tuple(bolds))
if len(fgraph) != bold.shape[1]:
raise Exception('Nb positions not consistent between BOLD (%d) '
'and mesh (%d)' % (bold.shape[1], len(fgraph)))
# discard bad data (bold with var=0 and nan values):
discard_bad_data(bold, mask, time_axis=0)
return mask, mask_meta_obj, mask_loaded_from_file, bold, session_scans, \
graphs, edge_lengths
def load_fmri_surf_data(boldFiles, meshFile, roiMaskFile=None):
""" Load FMRI BOLD surface data from files.
Arguments:
boldFiles -- pathes to 2D texture data files (ie multisession data).
roiMaskFile -- a path to a 1D texture data file.
meshFile -- a path to a mesh file
Return:
A tuple (graphs, bolds, sessionScans, roiMask, dataHeader)
graphs -- a dict {roiId:roiGraph}. Each roiGraph is a list of neighbours list
bold -- a dict {roiId:bold}. Each bold is a 2D numpy float array
with axes [time,position]
sessionScans -- a list of scan indexes for every session
roiMask -- a 1D numpy int array defining ROIs (0 stands for the background)
dataHeader -- the header from the BOLD data file
"""
#Load ROIs:
pyhrf.verbose(1, 'load roi mask: ' + roiMaskFile)
#roiMask = Texture.read(roiMaskFile).data.astype(int)
roiMask,_ = read_texture(roiMaskFile)
pyhrf.verbose(1, 'roi mask shape: ' + str(roiMask.shape))
#Load BOLD:
lastScan = 0
sessionScans = []
bolds = []
for boldFile in boldFiles:
pyhrf.verbose(1, 'load bold: ' + boldFile)
b,_ = read_texture(boldFile)
pyhrf.verbose(1, 'bold shape: ' + str(b.shape))
bolds.append(b)
sessionScans.append(np.arange(lastScan, lastScan+b.shape[0], dtype=int))
lastScan += b.shape[0]
bold = np.concatenate(tuple(bolds))
# discard bad data (bold with var=0 and nan values):
discard_bad_data(bold, roiMask, time_axis=0)
#Load mesh:
pyhrf.verbose(1, 'load mesh: ' + meshFile)
coords,triangles,coord_sys = read_mesh(meshFile)
#from soma import aims
# mesh = aims.read(meshFile)
# triangles = [t.arraydata() for t in mesh.polygon().list()]
pyhrf.verbose(3, 'building graph ... ')
wholeGraph = graph_from_mesh(triangles)
assert graph_is_sane(wholeGraph)
pyhrf.verbose(1, 'Mesh has %d nodes' %len(wholeGraph))
assert len(roiMask) == len(wholeGraph)
assert bold.shape[1] == len(wholeGraph)
pyhrf.verbose(3, 'Computing length of edges ... ')
edges_l = np.array([np.array([distance(coords[i],coords[n],coord_sys.xform) \
for n in nl]) \
for i,nl in enumerate(wholeGraph)], dtype=object)
#Split bold and graph into rois:
roiBold = {}
graphs = {}
edge_lengthes = {}
for roiId in np.unique(roiMask):
mroi = np.where(roiMask==roiId)
g, nm = sub_graph(wholeGraph, mroi[0])
edge_lengthes[roiId] = edges_l[mroi]
graphs[roiId] = g
roiBold[roiId] = bold[:, mroi[0]].astype(np.float32)
return graphs, roiBold, sessionScans, roiMask, edge_lengthes
def parcellation_ward_spatial(func_data, n_clusters, graph=None):
"""
Make parcellation based upon ward hierarchical clustering from scikit-learn
Inputs: - func_data: functional data: array of shape
(nb_positions, dim_feature_1, [dim_feature2, ...])
- n_clusters: chosen number of clusters to create
- graph: adjacency list defining neighbours
(if None, no connectivity defined: clustering is spatially
independent)
Output: parcellation labels
"""
from sklearn.cluster import Ward
from pyhrf.graph import graph_to_sparse_matrix, sub_graph
#from pyhrf.tools import cartesian
# print 'graph:'
# print graph
if graph is not None:
#print 'connectivity;'
#print connectivity.todense()
labels = np.zeros(len(graph), dtype=np.int32)
ccs = connected_components(graph)
n_tot = len(graph) * 1.
ncs = np.zeros(len(ccs), dtype=np.int32)
for icc,cc in enumerate(ccs):
nc = int(np.round(n_clusters * len(cc)/n_tot))
if nc == 0:
nc = 1
ncs[icc] = nc
max_nc = np.argmax(ncs)
ncs[max_nc] = n_clusters - sum(ncs[0:max_nc]) - sum(ncs[max_nc+1:])
assert sum(ncs) == n_clusters
assert (ncs > 0).all()
pyhrf.verbose(1, 'Found %d connected components (CC) of sizes: %s' \
%(len(ccs), ' ,'.join([str(len(cc)) for cc in ccs])))
pyhrf.verbose(1, 'Nb of clusters to search in each CC: %s' \
%' ,'.join(map(str,ncs)))
for nc,cc in zip(ncs,ccs):
#print 'cc:', len(cc)
#print 'cartesian:', list(cartesian(cc,cc))
if len(cc) < 2:
continue
if len(cc) < len(graph):
cc_graph,_ = sub_graph(graph, cc)
else:
cc_graph = graph
cc_connectivity = graph_to_sparse_matrix(cc_graph)
#print 'compute subslice ...'
# sub_slice = tuple(np.array(list(cartesian(cc,cc))).T) #indexes of the subpart of the connectivity matrix
#print 'sub_slice:'
#print sub_slice
#print 'subslice connectivity matrix ...'
#cc_connectivity = connectivity.tolil()[sub_slice].reshape((len(cc),len(cc))).tocoo() #indexes applied to the matrix connectivity (coo unsubscriptable)
#print 'cc_connectivity'
#print cc_connectivity.todense()
cc_data = func_data[cc]
pyhrf.verbose(2, 'Launch spatial Ward (nclusters=%d) '\
' on data of shape %s' %(nc, str(cc_data.shape)))
ward_object = Ward(n_clusters=nc, connectivity=cc_connectivity).fit(cc_data)
labels[cc] += ward_object.labels_ + 1 + labels.max()
else:
ward_object = Ward(n_clusters=n_clusters).fit(func_data) # connectivity=None
labels = ward_object.labels_ + 1
return labels
def parcellation_ward_spatial(func_data, n_clusters, graph=None):
"""
Make parcellation based upon ward hierarchical clustering from scikit-learn
Inputs: - func_data: functional data: array of shape
(nb_positions, dim_feature_1, [dim_feature2, ...])
- n_clusters: chosen number of clusters to create
- graph: adjacency list defining neighbours
(if None, no connectivity defined: clustering is spatially
independent)
Output: parcellation labels
"""
try:
# sklearn version < 0.17
from sklearn.cluster import Ward as AgglomerativeClustering
except ImportError:
from sklearn.cluster import AgglomerativeClustering
from pyhrf.graph import graph_to_sparse_matrix, sub_graph
if graph is not None:
labels = np.zeros(len(graph), dtype=np.int32)
ccs = connected_components(graph)
n_tot = len(graph) * 1.
ncs = np.zeros(len(ccs), dtype=np.int32)
for icc, cc in enumerate(ccs):
nc = int(np.round(n_clusters * len(cc) / n_tot))
if nc == 0:
nc = 1
ncs[icc] = nc
max_nc = np.argmax(ncs)
ncs[max_nc] = n_clusters - sum(ncs[0:max_nc]) - sum(ncs[max_nc + 1:])
assert sum(ncs) == n_clusters
assert (ncs > 0).all()
logger.info('Found %d connected components (CC) of sizes: %s', len(ccs),
' ,'.join([str(len(cc)) for cc in ccs]))
logger.info('Nb of clusters to search in each CC: %s',
' ,'.join(map(str, ncs)))
for nc, cc in zip(ncs, ccs):
if len(cc) < 2:
continue
if len(cc) < len(graph):
cc_graph, _ = sub_graph(graph, cc)
else:
cc_graph = graph
cc_connectivity = graph_to_sparse_matrix(cc_graph)
cc_data = func_data[cc]
logger.info('Launch spatial Ward (nclusters=%d) on data of shape %s',
nc, str(cc_data.shape))
ward_object = AgglomerativeClustering(
n_clusters=nc, connectivity=cc_connectivity
).fit(cc_data)
labels[cc] += ward_object.labels_ + 1 + labels.max()
else:
ward_object = AgglomerativeClustering(
n_clusters=n_clusters
).fit(func_data) # connectivity=None
labels = ward_object.labels_ + 1
return labels
def test_single_surface_PFPS_ML(self):
"""PF estimation method : path sampling. ML on p(label|beta).
topology from a surfacic RDI
"""
# generate a field:
beta = 0.4
nbClasses = 2
print 'generating potts ..., beta =', beta
# grab surfacic data:
from pyhrf.graph import graph_from_mesh, sub_graph, graph_is_sane
from pyhrf.tools._io.tio import Texture
from soma import aims
print 'import done'
roiId = 20
mfn = pyhrf.get_data_file_name('right_hemisphere.mesh')
print 'mesh file:', mfn
mesh = aims.read(mfn)
print 'mesh read'
triangles = [t.arraydata() for t in mesh.polygon().list()]
print 'building graph ... '
wholeGraph = graph_from_mesh(triangles)
roiMaskFile = pyhrf.get_data_file_name('roimask_gyrii_tuned.tex')
roiMask = Texture.read(roiMaskFile).data.astype(int)
mroi = np.where(roiMask==roiId)
g, nm = sub_graph(wholeGraph, mroi[0])
print "g:", len(g), len(g[0])
nnodes = len(g)
points = np.vstack([v.arraydata() for v in mesh.vertex().list()])
weights = [[1./dist(points[j],points[k]) for k in g[j]]
for j in xrange(nnodes)]
print "weights:", len(weights), len(weights[0])
if 1:
for j in xrange(nnodes):
s = sum(weights[j]) * 1.
for k in xrange(len(weights[j])):
weights[j][k] = weights[j][k]/s * len(weights[j])
labels = genPotts(g, beta, nbClasses, weights=weights)
print labels
# partition function estimation
gridLnz = Cpt_Vec_Estim_lnZ_Graph(g, nbClasses,
GraphWeight=weights)
print 'gridLnz with weights:'
print gridLnz
# beta estimation
be, pb = beta_estim_obs_field(g, labels, gridLnz, 'ML', weights)
print 'betaML:', be
weights = None
gridLnz = Cpt_Vec_Estim_lnZ_Graph(g, nbClasses,
GraphWeight=weights)
print 'gridLnz without weights:'
print gridLnz
# beta estimation
be, pb = beta_estim_obs_field(g, labels, gridLnz, 'ML', weights)
print 'betaML:', be
gridPace = gridLnz[1][1] - gridLnz[1][0]
assert abs(be-beta) <= gridPace
def load_surf_bold_mask(bold_files, mesh_file, mask_file=None):
from pyhrf.tools._io import read_mesh, read_texture, discard_bad_data
logger.info('Load mesh: ' + mesh_file)
coords, triangles, coord_sys = read_mesh(mesh_file)
logger.debug('Build graph...')
fgraph = graph_from_mesh(triangles)
assert graph_is_sane(fgraph)
logger.info('Mesh has %d nodes', len(fgraph))
logger.debug('Compute length of edges ... ')
edges_l = np.array([
np.array([distance(coords[i], coords[n], coord_sys.xform) for n in nl])
for i, nl in enumerate(fgraph)
],
dtype=object)
if mask_file is None or not op.exists(mask_file):
logger.warning('Mask file %s does not exist. Taking '
'all nodes ...', mask_file)
mask = np.ones(len(fgraph))
mask_meta_obj = None
mask_loaded_from_file = False
else:
mask, mask_meta_obj = read_texture(mask_file)
mask_loaded_from_file = True
if not (np.round(mask) == mask).all():
raise Exception("Mask is not n-ary")
if len(mask) != len(fgraph):
raise Exception('Size of mask (%d) is different from size '
'of graph (%d)' % (len(mask), len(fgraph)))
mask = mask.astype(np.int32)
if mask.min() == -1:
mask += 1
# Split graph into rois:
graphs = {}
edge_lengths = {}
for roi_id in np.unique(mask):
mroi = np.where(mask == roi_id)
graph, _ = sub_graph(fgraph, mroi[0])
edge_lengths[roi_id] = edges_l[mroi]
graphs[roi_id] = graph
# Load BOLD:
last_scan = 0
session_scans = []
bolds = []
for bold_file in bold_files:
logger.info('load bold: %s', bold_file)
bold, _ = read_texture(bold_file)
logger.info('bold shape: %s', str(bold.shape))
bolds.append(bold)
session_scans.append(
np.arange(last_scan, last_scan + bold.shape[0], dtype=int))
last_scan += bold.shape[0]
bold = np.concatenate(tuple(bolds))
if len(fgraph) != bold.shape[1]:
raise Exception('Nb positions not consistent between BOLD (%d) '
'and mesh (%d)' % (bold.shape[1], len(fgraph)))
# discard bad data (bold with var=0 and nan values):
discard_bad_data(bold, mask, time_axis=0)
return mask, mask_meta_obj, mask_loaded_from_file, bold, session_scans, \
graphs, edge_lengths
def parcellation_ward_spatial(func_data, n_clusters, graph=None):
"""
Make parcellation based upon ward hierarchical clustering from scikit-learn
Inputs: - func_data: functional data: array of shape
(nb_positions, dim_feature_1, [dim_feature2, ...])
- n_clusters: chosen number of clusters to create
- graph: adjacency list defining neighbours
(if None, no connectivity defined: clustering is spatially
independent)
Output: parcellation labels
"""
try:
# sklearn version < 0.17
from sklearn.cluster import Ward as AgglomerativeClustering
except ImportError:
from sklearn.cluster import AgglomerativeClustering
from pyhrf.graph import graph_to_sparse_matrix, sub_graph
if graph is not None:
labels = np.zeros(len(graph), dtype=np.int32)
ccs = connected_components(graph)
n_tot = len(graph) * 1.
ncs = np.zeros(len(ccs), dtype=np.int32)
for icc, cc in enumerate(ccs):
nc = int(np.round(n_clusters * len(cc) / n_tot))
if nc == 0:
nc = 1
ncs[icc] = nc
max_nc = np.argmax(ncs)
ncs[max_nc] = n_clusters - sum(ncs[0:max_nc]) - sum(ncs[max_nc + 1:])
assert sum(ncs) == n_clusters
assert (ncs > 0).all()
logger.info('Found %d connected components (CC) of sizes: %s',
len(ccs), ' ,'.join([str(len(cc)) for cc in ccs]))
logger.info('Nb of clusters to search in each CC: %s',
' ,'.join(map(str, ncs)))
for nc, cc in zip(ncs, ccs):
if len(cc) < 2:
continue
if len(cc) < len(graph):
cc_graph, _ = sub_graph(graph, cc)
else:
cc_graph = graph
cc_connectivity = graph_to_sparse_matrix(cc_graph)
cc_data = func_data[cc]
logger.info(
'Launch spatial Ward (nclusters=%d) on data of shape %s', nc,
str(cc_data.shape))
ward_object = AgglomerativeClustering(
n_clusters=nc, connectivity=cc_connectivity).fit(cc_data)
labels[cc] += ward_object.labels_ + 1 + labels.max()
else:
ward_object = AgglomerativeClustering(n_clusters=n_clusters).fit(
func_data) # connectivity=None
labels = ward_object.labels_ + 1
return labels