def plot_connectivity_glassbrain(subject_id, group, pc, roi_coords,
suffix, session='func1',
preprocessing_folder='pipeline_1',
save=True, msdl=False):
"""Plots connectome of pc
"""
title = '-'.join([suffix, group, subject_id, session])
output_folder = os.path.join(set_figure_base_dir('connectivity'), suffix)
if not os.path.isdir(output_folder):
os.makedirs(output_folder)
output_file = os.path.join(output_folder,
'_'.join([suffix, 'connectome', group,
session,
preprocessing_folder, subject_id]))
if msdl:
title += '_msdl'
output_file += '_msdl'
plt.figure(figsize=(10, 20), dpi=90)
if save:
plot_connectome(pc, roi_coords, edge_threshold='90%', title=title,
output_file=output_file)
else:
plot_connectome(pc, roi_coords, edge_threshold='90%', title=title)
def plot_connectivity_matrix(subject_id,
group,
pc,
roi_names,
suffix,
session='func1',
preprocessing_folder='pipeline_1',
save=True,
msdl=False):
"""Plots connectivity matrix of pc
"""
title = '-'.join([suffix, group, subject_id, session])
# plot matrix
output_folder = os.path.join(set_figure_base_dir('connectivity'), suffix)
if not os.path.isdir(output_folder):
os.makedirs(output_folder)
output_file = os.path.join(
output_folder, '_'.join([
suffix, 'matrix', group, session, preprocessing_folder, subject_id
]))
if msdl:
title += '_msdl'
output_file += '_msdl'
if msdl:
plt.figure(figsize=(12, 12))
else:
plt.figure(figsize=(8, 8))
plt.imshow(pc, cmap=cm.bwr, interpolation='nearest', vmin=-1, vmax=1)
plt.colorbar()
plt.xticks(range(len(roi_names)),
roi_names,
rotation='vertical',
fontsize=16)
plt.yticks(range(len(roi_names)), roi_names, fontsize=16)
plt.title(title, fontsize=20)
plt.tight_layout()
if save:
plt.savefig(output_file)
def plot_connectivity_matrix(subject_id, group, pc, roi_names,
suffix, session='func1',
preprocessing_folder='pipeline_1',
save=True, msdl=False):
"""Plots connectivity matrix of pc
"""
title = '-'.join([suffix, group, subject_id, session])
# plot matrix
output_folder = os.path.join(set_figure_base_dir('connectivity'), suffix)
if not os.path.isdir(output_folder):
os.makedirs(output_folder)
output_file = os.path.join(output_folder,
'_'.join([suffix, 'matrix', group,
session,
preprocessing_folder, subject_id]))
if msdl:
title += '_msdl'
output_file += '_msdl'
if msdl:
plt.figure(figsize=(12, 12))
else:
plt.figure(figsize=(8, 8))
plt.imshow(pc, cmap=cm.bwr, interpolation='nearest',
vmin=-1, vmax=1)
plt.colorbar()
plt.xticks(range(len(roi_names)), roi_names,
rotation='vertical', fontsize=16)
plt.yticks(range(len(roi_names)), roi_names, fontsize=16)
plt.title(title, fontsize=20)
plt.tight_layout()
if save:
plt.savefig(output_file)
def plot_connectivity_glassbrain(subject_id,
group,
pc,
roi_coords,
suffix,
session='func1',
preprocessing_folder='pipeline_1',
save=True,
msdl=False):
"""Plots connectome of pc
"""
title = '-'.join([suffix, group, subject_id, session])
output_folder = os.path.join(set_figure_base_dir('connectivity'), suffix)
if not os.path.isdir(output_folder):
os.makedirs(output_folder)
output_file = os.path.join(
output_folder, '_'.join([
suffix, 'connectome', group, session, preprocessing_folder,
subject_id
]))
if msdl:
title += '_msdl'
output_file += '_msdl'
plt.figure(figsize=(10, 20), dpi=90)
if save:
plot_connectome(pc,
roi_coords,
edge_threshold='90%',
title=title,
output_file=output_file)
else:
plot_connectome(pc, roi_coords, edge_threshold='90%', title=title)
def pairwise_classification(X, y, title=''):
""" Computes and plots accuracy of pairwise classification model
"""
# Ridge classification
rdgc = RidgeClassifierCV(alphas=np.logspace(-3, 3, 7))
# Support Vector classification
svc = LinearSVC(penalty='l1', dual=False)
# Linear Discriminant Analysis
lda = LDA()
# Logistic Regression
logit = LogisticRegression(penalty='l1', random_state=42)
estimator_str = ['svc', 'lda', 'rdgc', 'logit']
# train size
train_size = np.linspace(.2, .9, 8)
best_w = []
best_acc = 0
for e in estimator_str:
estimator = eval(e)
mean_acc = []
std_acc = []
for ts in train_size:
sss = StratifiedShuffleSplit(y, n_iter=50, train_size=ts,
random_state=42)
# Compute accuracies
accuracy = []
w = []
for train, test in sss:
estimator.fit(X[train], y[train])
accuracy.append(estimator.score(X[test], y[test]))
if e != 'rdgc' and e != 'lda':
w.append(estimator.coef_)
acc = np.mean(accuracy)
acc_std = np.std(accuracy)/2
mean_acc.append(acc)
std_acc.append(acc_std)
if len(w) > 0 and acc > best_acc :
best_acc = acc
best_w = np.mean(w, axis=0)
bl = plt.plot(train_size, mean_acc)
plt.fill_between(train_size,
np.sum([mean_acc, std_acc], axis=0),
np.subtract(mean_acc, std_acc),
facecolor=bl[0].get_c(),
alpha=0.1)
plt.xticks(fontsize=16)
plt.yticks(fontsize=16)
plt.legend(estimator_str, loc='best')
plt.xlabel('Train size', fontsize=16)
plt.ylabel('Accuracy', fontsize=16)
plt.ylim([.3, .9])
plt.grid()
plt.title('Classification ' + title, fontsize=16)
if msdl:
msdl_str = 'msdl'
else:
msdl_str = 'rois'
output_folder = os.path.join(set_figure_base_dir('classification'),
metric, session, msdl_str)
if not os.path.isdir(output_folder):
os.makedirs(output_folder)
output_file = os.path.join(output_folder, 'accuracy_' + title)
plt.savefig(output_file)
return best_w, best_acc
gr_j = dataset.group_indices[groups[j]]
Xp = np.vstack((X[gr_i, :], X[gr_j, :]))
yp = np.array([0] * len(gr_i) + [1] * len(gr_j))
output = '_'.join([groups[i], groups[j], metric, session,
msdl_str, preprocessing_folder])
plt.figure()
w,a = pairwise_classification(Xp, yp, title=output)
print groups[i], groups[j], a
t = np.zeros((len(roi_names), len(roi_names)))
t[ind] = np.abs(w)
t = (t + t.T) / 2.
if msdl:
msdl_str = 'msdl'
else:
msdl_str = 'rois'
output_folder = os.path.join(set_figure_base_dir('classification'),
metric, session, msdl_str)
if not os.path.isdir(output_folder):
os.makedirs(output_folder)
output_file = os.path.join(output_folder, 'connectome_' + output)
plot_connectome(t, roi_coords, title=output,
output_file=output_file,
annotate=True,
edge_threshold='0%')
# 1 vs rest
for i in range(3):
gr_i = dataset.group_indices[groups[i]]
yr = np.zeros(X.shape[0])
yr[gr_i] = 1
do_cca=True,
threshold=3.,
n_init=10,
standardize=True,
random_state=42,
n_jobs=2,
verbose=2)
CANICA_PATH = set_data_base_dir('Dynacomp/canica')
output_file = os.path.join(CANICA_PATH,
'canica_' + str(n_components) + '.nii.gz')
if not os.path.isfile(output_file):
canica.fit(func)
components_img = canica.masker_.inverse_transform(canica.components_)
components_img.to_filename(output_file)
else:
components_img = output_file
FIG_PATH = set_figure_base_dir('canica')
for i, cur_img in enumerate(iter_img(components_img)):
output_file = os.path.join(FIG_PATH,
'canica_' + str(n_components) + '_IC%d' % i)
plot_stat_map(cur_img,
display_mode="z",
title="IC %d" % i,
cut_coords=1,
colorbar=False,
output_file=output_file)
# -*- coding: utf-8 -*-
"""
Created on Wed Apr 1 09:16:38 2015
@author: mr243268
"""
import os
from loader import load_dynacomp, set_figure_base_dir
from nilearn.plotting import plot_roi
dataset = load_dynacomp()
FIG_DIR = set_figure_base_dir('rois')
for i in range(len(dataset.subjects)):
for k in sorted(dataset.rois[i].keys()):
output_file = os.path.join(FIG_DIR, k)
plot_roi(dataset.rois[i][k], title=k, output_file=output_file)
break
from nilearn.image import iter_img
dataset = load_dynacomp()
func = dataset.func1
n_components = 20
canica = CanICA(n_components=n_components, mask=dataset.mask,
smoothing_fwhm=None, do_cca=True, threshold=3.,
n_init=10, standardize=True, random_state=42,
n_jobs=2, verbose=2)
CANICA_PATH = set_data_base_dir('Dynacomp/canica')
output_file = os.path.join(CANICA_PATH,
'canica_' + str(n_components) + '.nii.gz')
if not os.path.isfile(output_file):
canica.fit(func)
components_img = canica.masker_.inverse_transform(canica.components_)
components_img.to_filename(output_file)
else:
components_img = output_file
FIG_PATH = set_figure_base_dir('canica')
for i, cur_img in enumerate(iter_img(components_img)):
output_file = os.path.join(FIG_PATH,
'canica_'+ str(n_components) +'_IC%d' % i )
plot_stat_map(cur_img, display_mode="z", title="IC %d" % i, cut_coords=1,
colorbar=False, output_file=output_file)
