def plotScalpmapsOddball(epo, epo_t, clab, mrk_class, intervals, mnt):
lim_min, lim_max = 100, -100
for mrk_cl in np.unique(mrk_class):
epo_cls = np.mean(epo[:, :, mrk_class == mrk_cl], axis=2)
for ival in intervals:
idx_range = np.where((epo_t >= ival[0]) & (epo_t <= ival[1]))
epo_cls_ival = np.mean(epo_cls[idx_range, :], axis=0)
lim_min = np.min(
epo_cls_ival) if np.min(epo_cls_ival) < lim_min else lim_min
lim_max = np.max(
epo_cls_ival) if np.max(epo_cls_ival) > lim_max else lim_max
for mrk_cl in np.unique(mrk_class):
label = 'Target' if mrk_cl == 0 else 'Non-Target'
epo_cls = np.mean(epo[:, :, mrk_class == mrk_cl], axis=2)
print(epo_cls.shape)
for ival in intervals:
plt.figure(figsize=(6, 6))
idx_range = np.where((epo_t >= ival[0]) & (epo_t <= ival[1]))
print(epo_cls[idx_range, :].shape)
epo_cls_ival = np.mean(epo_cls[idx_range, :].squeeze(), axis=0)
print(epo_cls_ival.shape)
plt.title(''.join(
[label, ', Interval ',
str(ival[0]), ' - ',
str(ival[1])]))
bci.scalpmap(mnt,
epo_cls_ival,
clim=[lim_min, lim_max],
cb_label='µV',
clab=clab)
def performPCA(X,mnt=None):
Xmean = np.mean(X,axis=1)
Xm = X-Xmean[:,np.newaxis]
d,v=np.linalg.eig(np.cov(X))
print(d)
for i in range(v.shape[0]):
plt.figure()
bci.scalpmap(mnt,v[:,i])
return d,v
def plotPrincipalComponents(cnt, mnt, clab, n=3):
X = cnt
Xmean = np.mean(X, axis=1)
X = X - Xmean[:, np.newaxis]
d, v = np.linalg.eig(np.cov(X))
for i in range(n):
plt.figure(figsize=(6, 6))
plt.title(''.join(
['Component #',
str(i + 1), ', std:',
str(np.sqrt(d[i]))]))
bci.scalpmap(mnt, v[:, i], clim='sym', cb_label='[a.u.]', clab=clab)
return d, v
def plotScalpmapsArtifact(epo, epo_t, clab, mrk_class, intervals, mnt):
fig, axs = plt.subplots(len(np.unique(mrk_class)),
len(intervals) + 1,
figsize=(24, 8))
for i, mrk_cl in zip(range(len(np.unique(mrk_class))),
np.unique(mrk_class)):
epo_cls = np.mean(epo[:, :, mrk_class == mrk_cl], axis=2)
lim_min, lim_max = 0, 0
print(artifactDict[mrk_cl])
for j, ival in zip(range(len(intervals)), intervals):
idx_range = np.where((epo_t >= ival[0]) & (epo_t <= ival[1]))
epo_cls_ival = np.mean(epo_cls[idx_range, :].squeeze(), axis=0)
lim_min = np.min(
epo_cls_ival) if np.min(epo_cls_ival) < lim_min else lim_min
lim_max = np.max(
epo_cls_ival) if np.max(epo_cls_ival) > lim_max else lim_max
#axs[i,j].title(''.join(['Interval',str(ival[0]),' - ',str(ival[1])]))
pcm = bci.scalpmap(mnt,
epo_cls_ival,
clim=[lim_min, lim_max],
cb_label='µV',
ax=axs[i, j])
axs[i, len(intervals)].axis('off')
fig.colorbar(pcm, shrink=.6, ax=axs[i, len(intervals)])
fig.tight_layout(pad=0.0)
plt.show()
# Get the indexes of the 6 maximum eigenvalues
eigidx = np.argsort(ddif)[-numCSP:]
# Spatial filtering of signals
CSPs = apply_spatial_filter(epo, W[:, eigidx])
AllCSP = apply_spatial_filter(epo, W)
A = np.dot(np.dot(np.cov(epo.reshape(epo.shape[1], -1)), W),
np.linalg.inv(np.cov(AllCSP.reshape(epo_o.shape[1], -1))))
plt.figure(figsize=(8, 8))
n = 221
for i in range(eigidx.shape[0]):
plt.subplot(n + i)
plt.title('CSP filter ' + str(4 - i))
bci.scalpmap(mnt, A[:, eigidx[i]], clim='sym', cb_label='[a.u.]')
epoCSP = np.abs(signal.hilbert(CSPs, axis=0))
aveCSPsT = np.mean(epoCSP[:, :, mrk_class == 0], axis=2)
aveCSPsArtifact = np.mean(epoCSP[:, :, mrk_class == 1], axis=2)
plt.figure(figsize=(10, 8))
n = 221
for i in range(eigidx.shape[0]):
plt.subplot(n + i)
plt.title('Averaged ERD, EigNum = ' + str(eigidx[i]))
plt.plot(epo_t_o, aveCSPsT[:, i], label='target')
plt.plot(epo_t_o,
aveCSPsArtifact[:, i],
label=str(artifactDict[artifact]))