def _compute_surrogate_density_dev(BFLs,xyz,dmax,nsamples=1):
"""
caveat: does not work for nsamples>1
This function computes a surrogate density
using graph diffusion techniques
"""
import nipy.neurospin.utils.smoothing.smoothing as smoothing
nvox = xyz.shape[0]
nbsubj = np.size(BFLs)
nlm = np.array([BFLs[s].k for s in range(nbsubj)])
aux = np.zeros((nvox, nsamples*nbsubj))
for s in range(nbsubj):
if nlm[s]>0:
for it in range(nsamples):
js = (nvox*nr.rand(nlm[s])).astype(np.int)
aux[js,s*nsamples+it] = 1.
aux = smoothing.cartesian_smoothing(xyz.T,aux,dmax)
surweight = np.zeros((nvox*nsamples,nbsubj),'d')
for s in range(nbsubj):
surweight[:,s] = np.reshape(aux[:,s*nsamples:(s+1)*nsamples],
(nvox*nsamples))
surweight = surweight*(2*np.pi*dmax*dmax)**1.5
return surweight
def _compute_density_dev(BFLs,xyz,dmax):
"""
Computation of the density of the BFLs points in the xyz volume
dmax is a scale parameter
"""
import nipy.neurospin.utils.smoothing.smoothing as smoothing
nvox = xyz.shape[0]
nbsubj = np.size(BFLs)
weight = np.zeros((nvox,nbsubj),'d')
nlm = np.array([BFLs[s].k for s in range(nbsubj)])
for s in range(nbsubj):
if nlm[s]>0:
weight[BFLs[s].seed,s] = 1
weight = smoothing.cartesian_smoothing(np.transpose(xyz),weight,dmax)
weight = weight*(2*np.pi*dmax*dmax)**1.5
return weight