def stats_save_nii(corrs,
filename,
affine,
corr_mask=get_HOcort(),
size=[60, 60, 60],
ksize=[3, 3, 3],
strides=[1, 1, 1],
p=0.05,
df=20,
correct_method=None,
smooth=False,
plotrlt=True,
img_background=None):
"""
Save the searchlight RSA statistical results as a NIfTI file for fMRI
Parameters
----------
corrs : array
The statistical results between behavioral data and fMRI data for searchlight.
The shape of RDMs is [n_x, n_y, n_z, 2]. n_x, n_y, n_z represent the number of calculation units for searchlight
along the x, y, z axis and 2 represents a t-value and a p-value.
filename : string. Default is 'rsa_result.nii'.
The file path+filename for the result .nii file.
If the filename does not end in ".nii", it will be filled in automatically.
affine : array or list
The position information of the fMRI-image array data in a reference space.
corr_mask : string
The filename of a mask data for correcting the RSA result.
It can just be one of your fMRI data files in your experiment for a mask file for ROI. If the corr_mask is a
filename of a ROI mask file, only the RSA results in ROI will be visible.
size : array or list [nx, ny, nz]. Default is [60, 60, 60].
The size of the fMRI-img in your experiments.
ksize : array or list [kx, ky, kz]. Default is [3, 3, 3].
The size of the fMRI-img. nx, ny, nz represent the number of voxels along the x, y, z axis
strides : array or list [sx, sy, sz]. Default is [1, 1, 1].
The strides for calculating along the x, y, z axis.
p : float. Default is 0.05.
The threshold of p-values.
Only the results those p-values are lower than this value will be visible.
df : int. Default is 20.
The degree of freedom.
correct_method : None or string 'FWE' or 'FDR'. Default is None.
The method for correcting the RSA results.
If correct_method='FWE', here the FWE-correction will be used. If correct_methd='FDR', here the FDR-correction
will be used. If correct_method=None, no correction.
Only when p<1, correct_method works.
smooth : bool True or False. Default is False.
Smooth the RSA result or not.
plotrlt : bool True or False. Default is True.
Plot the RSA result automatically or not.
img_background : None or string. Default if None.
The filename of a background image that the RSA results will be plotted on the top of it.
If img_background=None, the background will be ch2.nii.gz.
Only when plotrlt=True, img_background works.
Returns
-------
img : array
The array of the statistical results t-values map.
The shape is [nx, ny, nz]. nx, ny, nz represent the size of the fMRI-img.
Notes
-----
A result .nii file of searchlight statistical results will be generated at the corresponding address of filename.
"""
# get the size of the fMRI-img
nx = size[0]
ny = size[1]
nz = size[2]
# the size of the calculation units for searchlight
kx = ksize[0]
ky = ksize[1]
kz = ksize[2]
# strides for calculating along the x, y, z axis
sx = strides[0]
sy = strides[1]
sz = strides[2]
# calculate the number of the calculation units in the x, y, z directions
n_x = np.shape(corrs)[0]
n_y = np.shape(corrs)[1]
n_z = np.shape(corrs)[2]
# initialize the indexes to record the number of valid values for each voxel
index = np.zeros([nx, ny, nz], dtype=np.int)
# initialize the img array to save the sum-r-value for each voxel
img_nii = np.zeros([nx, ny, nz], dtype=np.float64)
# iterate through all the calculation units
# calculate the indexs
for i in range(n_x):
for j in range(n_y):
for k in range(n_z):
x = i * sx
y = j * sy
z = k * sz
if (math.isnan(corrs[i, j, k, 0]) is False):
for k1 in range(kx):
for k2 in range(ky):
for k3 in range(kz):
index[x + k1, y + k2, z +
k3] = index[x + k1, y + k2, z + k3] + 1
# initialize a mask in order to record valid voxels (have qualified results)
mask = np.zeros([nx, ny, nz], dtype=np.int)
# get the p-values
corrsp = corrs[:, :, :, 1]
# calculate the number of voxels for correction
fadeimg = np.zeros([nx, ny, nz], dtype=np.int)
for i in range(n_x):
for j in range(n_y):
for k in range(n_z):
x = i * sx
y = j * sy
z = k * sz
# p-values<threshold-p
if corrsp[i, j, k] < p:
for k1 in range(kx):
for k2 in range(ky):
for k3 in range(kz):
fadeimg[x + k1, y + k2, z + k3] = 1
n_corrected = 0
for i in range(nx):
for j in range(ny):
for k in range(nz):
if fadeimg[i, j, k] == 1:
n_corrected = n_corrected + 1
print(str(n_corrected) + " voxels corrected")
# do the correction
if p < 1:
# FDR-correction
if correct_method == "FDR":
corrsp = fdr_correct(corrsp, p_threshold=p)
# FWE-correction
if correct_method == "FWE":
corrsp = fwe_correct(corrsp, p_threshold=p)
# iterate through all the calculation units again
# record the valid voxels
# [n_x, n_y, n_z] expanses into [nx, ny, nz] based on ksize & strides
for i in range(n_x):
for j in range(n_y):
for k in range(n_z):
x = i * sx
y = j * sy
z = k * sz
# p-values<threshold-p
if corrsp[i, j, k] < p:
for k1 in range(kx):
for k2 in range(ky):
for k3 in range(kz):
mask[x + k1, y + k2, z + k3] = 1
# sum of valid values in each calculation units
for i in range(n_x):
for j in range(n_y):
for k in range(n_z):
x = i * sx
y = j * sy
z = k * sz
if (math.isnan(corrs[i, j, k, 0]) is False):
for k1 in range(kx):
for k2 in range(ky):
for k3 in range(kz):
img_nii[x + k1, y + k2, z +
k3] = img_nii[x + k1, y + k2, z +
k3] + corrs[i, j, k, 0]
# initialize the newimg array to calculate the avg-r-value for each voxel
newimg_nii = np.full([nx, ny, nz], np.nan)
t_threshold = t.isf(p, df)
print(t_threshold)
# calculate the avg values of each valid voxel
for i in range(nx):
for j in range(ny):
for k in range(nz):
# valid voxel
if mask[i, j, k] == 1:
# sum-r-value/index
newimg_nii[i, j,
k] = float(img_nii[i, j, k] / index[i, j, k])
# set filename for result .nii file
if filename == None:
filename = "rsa_result.nii"
else:
q = ".nii" in filename
if q == True:
filename = filename
else:
filename = filename + ".nii"
# corr_mask != None
# use the mask file to correct RSA results
# in order to avoid results showing outside of the brain
if corr_mask == get_HOcort():
mask_to(get_bg_ch2bet(), filename, size, affine)
mask = nib.load(filename).get_data()
print(mask.shape)
else:
# load the array data of the mask file
mask = nib.load(corr_mask).get_data()
# do correction by the mask
for i in range(nx):
for j in range(ny):
for k in range(nz):
if (math.isnan(mask[i, j, k]) == True) or mask[i, j,
k] == 0:
newimg_nii[i, j, k] = np.nan
# do correction by the mask
for i in range(nx):
for j in range(ny):
for k in range(nz):
if (math.isnan(mask[i, j, k]) == True) or mask[i, j, k] == 0:
newimg_nii[i, j, k] = np.nan
if newimg_nii[i, j, k] < t_threshold and newimg_nii[i, j,
k] > 0:
newimg_nii[i, j, k] = np.nan
if newimg_nii[i, j, k] > -t_threshold and newimg_nii[i, j,
k] < 0:
newimg_nii[i, j, k] = np.nan
print(filename)
# save the .nii file for RSA results
file = nib.Nifti1Image(newimg_nii, affine)
if smooth == True:
# smooth the img data of the .nii file
file = smooth_img(file, fwhm='fast')
# save the result
nib.save(file, filename)
# determine if it has results
norlt = np.isnan(newimg_nii).all()
if norlt == True:
print("No RSA result.")
# determine plot the results or not
if norlt == False and plotrlt == True:
plot_brainrsa_rlts(filename, background=img_background, type='t')
print("File(" + filename + ") saves successfully!")
return newimg_nii
def test_fwe_correct(self):
p = np.random.rand(20, 22, 21)
p_threshold = 0.05
output = fwe_correct(p, p_threshold)
self.assertEqual(output.shape[0], 20)
def corr_save_nii(corrs,
filename,
affine,
size=[60, 60, 60],
ksize=[3, 3, 3],
strides=[1, 1, 1],
p=1,
r=0,
similarity=0,
distance=0,
correct_method=None,
correct_n=27):
nx = size[0]
ny = size[1]
nz = size[2]
kx = ksize[0]
ky = ksize[1]
kz = ksize[2]
sx = strides[0]
sy = strides[1]
sz = strides[2]
n_x = np.shape(corrs)[0]
n_y = np.shape(corrs)[1]
n_z = np.shape(corrs)[2]
index = np.zeros([nx, ny, nz], dtype=np.int)
img_nii = np.zeros([nx, ny, nz], dtype=np.float64)
nfdr = 0
for i in range(n_x):
for j in range(n_y):
for k in range(n_z):
x = i * sx
y = j * sy
z = k * sz
if (math.isnan(corrs[i, j, k, 0]) is False):
nfdr = nfdr + 1
for k1 in range(kx):
for k2 in range(ky):
for k3 in range(kz):
index[x + k1, y + k2, z +
k3] = index[x + k1, y + k2, z + k3] + 1
mask = np.zeros([nx, ny, nz], dtype=np.int)
corrsp = corrs[:, :, :, 1]
if p < 1:
if correct_method == "FDR":
corrsp = fdr_correct(corrsp, size=size, n=correct_n)
if correct_method == "FWE":
corrsp = fwe_correct(corrsp, size=size, n=correct_n)
for i in range(n_x):
for j in range(n_y):
for k in range(n_z):
x = i * sx
y = j * sy
z = k * sz
if (corrsp[i, j, k] < p) and (corrs[i, j, k, 0] > np.max(
[r, similarity, distance])):
for k1 in range(kx):
for k2 in range(ky):
for k3 in range(kz):
mask[x + k1, y + k2, z + k3] = 1
for i in range(n_x):
for j in range(n_y):
for k in range(n_z):
x = i * sx
y = j * sy
z = k * sz
if (math.isnan(corrs[i, j, k, 0]) is False):
for k1 in range(kx):
for k2 in range(ky):
for k3 in range(kz):
img_nii[x + k1, y + k2, z +
k3] = img_nii[x + k1, y + k2, z +
k3] + corrs[i, j, k, 0]
newimg_nii = np.full([nx, ny, nz], np.nan)
for i in range(nx):
for j in range(ny):
for k in range(nz):
if mask[i, j, k] == 1:
newimg_nii[i, j,
k] = float(img_nii[i, j, k] / index[i, j, k])
filename = filename + ".nii"
print(filename)
file = nib.Nifti1Image(newimg_nii, affine)
nib.save(file, filename)
print("File(" + filename + ") saves successfully!")
return newimg_nii
def stats_save_nii(stats, affine, filename=None, corr_mask=get_HOcort(), size=[60, 60, 60], ksize=[3, 3, 3],
strides=[1, 1, 1], p=0.05, correct_method=None, clusterp=0.05, smooth=False, plotrlt=True,
img_background=None):
"""
Save the searchlight RSA statistical results as a NIfTI file for fMRI
Parameters
----------
stats : array
The statistical results between behavioral data and fMRI data for searchlight.
The shape of RDMs is [n_x, n_y, n_z, 2]. n_x, n_y, n_z represent the number of calculation units for searchlight
along the x, y, z axis and 2 represents a t-value and a p-value.
If the filename does not end in ".nii", it will be filled in automatically.
affine : array or list
The position information of the fMRI-image array data in a reference space.
filename : string. Default is None - 'rsa_result.nii'.
The file path+filename for the result .nii file.
corr_mask : string
The filename of a mask data for correcting the RSA result.
It can just be one of your fMRI data files in your experiment for a mask file for ROI. If the corr_mask is a
filename of a ROI mask file, only the RSA results in ROI will be visible.
size : array or list [nx, ny, nz]. Default is [60, 60, 60].
The size of the fMRI-img in your experiments.
ksize : array or list [kx, ky, kz]. Default is [3, 3, 3].
The size of the calculation unit for searchlight.
kx, ky, kz represent the number of voxels along the x, y, z axis.
strides : array or list [sx, sy, sz]. Default is [1, 1, 1].
The strides for calculating along the x, y, z axis.
p : float. Default is 0.05.
The threshold of p-values.
Only the results those p-values are lower than this value will be visible.
correct_method : None or string 'FWE' or 'FDR' or 'Cluster-FWE' or 'Cluster-FDR'. Default is None.
The method for correcting the RSA results.
If correct_method='FWE', here the FWE-correction will be used. If correct_methd='FDR', here the FDR-correction
will be used. If correct_method='Cluster-FWE', here the Cluster-wise FWE-correction will be used. If
correct_methd='Cluster-FDR', here the Cluster-wise FDR-correction will be used. If correct_method=None, no
correction.
Only when p<1, correct_method works.
clusterp : float. Default is 0.05.
The threshold of p-value for cluster-wise correction.
Only when correct_method='Cluster-FDR' or 'Cluster-FWE', clusterp works.
smooth : bool True or False. Default is False.
Smooth the RSA result or not.
plotrlt : bool True or False. Default is True.
Plot the RSA result automatically or not.
img_background : None or string. Default if None.
The filename of a background image that the RSA results will be plotted on the top of it.
If img_background=None, the background will be ch2.nii.gz.
Only when plotrlt=True, img_background works.
Returns
-------
img : array
The array of the statistical results t-values map.
The shape is [nx, ny, nz]. nx, ny, nz represent the size of the fMRI-img.
Notes
-----
A result .nii file of searchlight statistical results will be generated at the corresponding address of filename.
"""
if len(np.shape(stats)) != 4 or len(np.shape(affine)) != 2 or np.shape(affine)[0] != 4 or np.shape(affine)[1] != 4:
return "Invalid input!"
# get the size of the fMRI-img
nx = size[0]
ny = size[1]
nz = size[2]
# the size of the calculation units for searchlight
kx = ksize[0]
ky = ksize[1]
kz = ksize[2]
rx = int((kx-1)/2)
ry = int((ky-1)/2)
rz = int((kz-1)/2)
# strides for calculating along the x, y, z axis
sx = strides[0]
sy = strides[1]
sz = strides[2]
# calculate the number of the calculation units in the x, y, z directions
n_x = np.shape(stats)[0]
n_y = np.shape(stats)[1]
n_z = np.shape(stats)[2]
img_nii = np.zeros([nx, ny, nz], dtype=np.float64)
# initialize a mask in order to record valid voxels (have qualified results)
mask = np.zeros([nx, ny, nz], dtype=np.int)
# get the p-values
statsp = stats[:, :, :, 1]
statst = stats[:, :, :, 0]
# calculate the number of voxels for correction
fadeimg = np.zeros([nx, ny, nz], dtype=np.int)
# iterate through all the calculation units
# calculate the indexs
for i in range(n_x):
for j in range(n_y):
for k in range(n_z):
x = i*sx
y = j*sy
z = k*sz
if statsp[i, j, k] < 1:
img_nii[x + rx, y + ry, z + rz] = statst[i, j, k]
if statsp[i, j, k] < p:
fadeimg[x + rx, y + ry, z + rz] = 1
n_corrected = 0
for i in range(nx):
for j in range(ny):
for k in range(nz):
if fadeimg[i, j, k] == 1:
n_corrected = n_corrected + 1
print(str(n_corrected)+" voxels will be corrected.")
# do the correction
if p < 1:
# FDR-correction
if correct_method == "FDR":
statsp = fdr_correct(statsp, p_threshold=p)
# FWE-correction
if correct_method == "FWE":
statsp = fwe_correct(statsp, p_threshold=p)
# Cluster-wise FDR-correction
if correct_method == "Cluster-FDR":
statsp = cluster_fdr_correct(statsp, p_threshold1=p, p_threshold2=clusterp)
# Cluster-wise FWE-correction
if correct_method == "Cluster-FWE":
statsp = cluster_fwe_correct(statsp, p_threshold1=p, p_threshold2=clusterp)
# iterate through all the calculation units again
print("Record the valid voxels.")
for i in range(n_x):
for j in range(n_y):
for k in range(n_z):
x = i * sx
y = j * sy
z = k * sz
if statsp[i, j, k] < p:
mask[x + rx, y + ry, z + rz] = 1
# initialize the newimg array to calculate the avg-r-value for each voxel
newimg_nii = np.full([nx, ny, nz], np.nan)
# set filename for result .nii file
if filename == None:
filename = "rsa_result.nii"
else:
q = ".nii" in filename
if q == True:
filename = filename
else:
filename = filename + ".nii"
# corr_mask != None
# use the mask file to correct RSA results
# in order to avoid results showing outside of the brain
if corr_mask == get_HOcort():
mask_to(get_bg_ch2bet(), size, affine, filename)
cmask = nib.load(filename).get_fdata()
else:
# load the array data of the mask file
cmask = nib.load(corr_mask).get_fdata()
# calculate the avg values of each valid voxel
for i in range(nx):
for j in range(ny):
for k in range(nz):
# valid voxel
if (math.isnan(cmask[i, j, k]) == False) and cmask[i, j, k] != 0 and mask[i, j, k] == 1:
# sum-r-value/index
newimg_nii[i, j, k] = img_nii[i, j, k]
print("Get RSA results.")
print(filename)
print("Save RSA results.")
# save the .nii file for RSA results
file = nib.Nifti1Image(newimg_nii, affine)
if smooth == True:
print("Smooth the results.")
# smooth the img data of the .nii file
file = smooth_img(file, fwhm='fast')
# save the result
nib.save(file, filename)
# determine if it has results
norlt = np.isnan(newimg_nii).all()
if norlt == True:
print("No RSA results.")
print("File("+filename+") saves successfully!")
# determine plot the results or not
if norlt == False and plotrlt == True:
print("Plot RSA results.")
plot_brainrsa_rlts(filename, background=img_background, type='t')
return newimg_nii
