def load_stim_file(subject, args):
stim_fname = op.join(
MMVT_DIR, subject, 'electrodes',
'{}{}.npz'.format(args.file_frefix, args.stim_channel))
stim = np.load(stim_fname)
labels, psd, time, freqs = (stim[k]
for k in ['labels', 'psd', 'time', 'freqs'])
bipolar = '-' in labels[0]
data = None
freqs_dim = psd.shape.index(len(freqs))
labels_dim = psd.shape.index(len(labels))
if time.ndim > 0:
time_dim = psd.shape.index(len(time))
else:
time_dim = next(iter(set(range(3)) - set([freqs_dim, labels_dim])))
T, L, F = psd.shape[time_dim], psd.shape[labels_dim], psd.shape[freqs_dim]
if args.downsample > 1:
time = utils.downsample(time, args.downsample)
colors = None
conditions = []
for freq_ind, (freq_from, freq_to) in enumerate(freqs):
if data is None:
# initialize the data matrix (electrodes_num x T x freqs_num)
data = np.zeros((L, T, F))
psd_slice = get_psd_freq_slice(psd, freq_ind, freqs_dim, time_dim)
if args.downsample > 1:
psd_slice = utils.downsample(psd_slice, args.downsample)
data[:, :, freq_ind] = psd_slice
data_min, data_max = utils.calc_min_max(psd_slice,
norm_percs=args.norm_percs)
if colors is None:
colors = np.zeros((*data.shape, 3))
for elec_ind, elec_name in enumerate(labels):
elec_group = utils.elec_group(elec_name, bipolar)
# if elec_group in ['LVF', 'RMT']:
# colors[elec_ind, :, freq_ind] = utils.mat_to_colors(psd_slice[elec_ind], data_min, data_max, colorsMap='BuGn')
# else:
colors[elec_ind, :,
freq_ind] = utils.mat_to_colors(psd_slice[elec_ind],
data_min,
data_max,
colorsMap=args.colors_map)
conditions.append('{}-{}Hz'.format(freq_from, freq_to))
output_fname = op.join(
MMVT_DIR, subject, 'electrodes',
'stim_electrodes_{}{}_{}.npz'.format(args.file_frefix,
'bipolar' if bipolar else '',
args.stim_channel))
print('Saving {}'.format(output_fname))
np.savez(output_fname,
data=data,
names=labels,
conditions=conditions,
colors=colors)
return dict(data=data, labels=labels, conditions=conditions, colors=colors)
def create_mmvt_file(subject,
stim_channel,
channel_list,
fs,
time,
stim,
theta,
downsample_ratio=10):
theta -= np.min(theta)
theta /= np.max(theta)
theta *= -1
stim = utils.downsample(stim.squeeze(), downsample_ratio)
stim[stim >= 0.5] = 1
stim[stim < 0.5] = 0
theta = utils.downsample_2d(theta.T, downsample_ratio)
data = np.vstack((stim, theta))
data = data[:, :, np.newaxis]
channel_list = [stim_channel] + channel_list
meta_data_fname = op.join(MMVT_DIR, subject, 'electrodes',
'electrodes_bipolar_data.npz')
np.savez(meta_data_fname,
data=data,
names=channel_list,
conditions=['on'],
dt=1 / fs)
def load_stim_file(subject, args):
stim_fname = op.join(MMVT_DIR, subject, 'electrodes', '{}{}.npz'.format(
args.file_frefix, args.stim_channel))
stim = np.load(stim_fname)
labels, psd, time, freqs = (stim[k] for k in ['labels', 'psd', 'time', 'freqs'])
bipolar = '-' in labels[0]
data = None
freqs_dim = psd.shape.index(len(freqs))
labels_dim = psd.shape.index(len(labels))
if time.ndim > 0:
time_dim = psd.shape.index(len(time))
else:
time_dim = next(iter(set(range(3)) - set([freqs_dim, labels_dim])))
T, L, F = psd.shape[time_dim], psd.shape[labels_dim], psd.shape[freqs_dim]
if args.downsample > 1:
time = utils.downsample(time, args.downsample)
colors = None
conditions = []
for freq_ind, (freq_from, freq_to) in enumerate(freqs):
if data is None:
# initialize the data matrix (electrodes_num x T x freqs_num)
data = np.zeros((L, T, F))
psd_slice = get_psd_freq_slice(psd, freq_ind, freqs_dim, time_dim)
if args.downsample > 1:
psd_slice = utils.downsample(psd_slice, args.downsample)
data[:, :, freq_ind] = psd_slice
data_min, data_max = utils.calc_min_max(psd_slice, norm_percs=args.norm_percs)
if colors is None:
colors = np.zeros((*data.shape, 3))
for elec_ind, elec_name in enumerate(labels):
elec_group = utils.elec_group(elec_name, bipolar)
# if elec_group in ['LVF', 'RMT']:
# colors[elec_ind, :, freq_ind] = utils.mat_to_colors(psd_slice[elec_ind], data_min, data_max, colorsMap='BuGn')
# else:
colors[elec_ind, :, freq_ind] = utils.mat_to_colors(psd_slice[elec_ind], data_min, data_max, colorsMap=args.colors_map)
conditions.append('{}-{}Hz'.format(freq_from, freq_to))
output_fname = op.join(MMVT_DIR, subject, 'electrodes', 'stim_electrodes_{}{}_{}.npz'.format(
args.file_frefix, 'bipolar' if bipolar else '', args.stim_channel))
print('Saving {}'.format(output_fname))
np.savez(output_fname, data=data, names=labels, conditions=conditions, colors=colors)
return dict(data=data, labels=labels, conditions=conditions, colors=colors)
def create_mmvt_file(subject,
stim_channel,
channel_list,
fs,
time,
stim,
theta,
downsample_ratio=10,
downsample_using_mean=True,
smooth_win_len=101):
theta = utils.downsample_2d(theta.T, downsample_ratio,
downsample_using_mean)
theta_smooth = np.zeros((theta.shape[0], theta.shape[1]))
for k in range(theta.shape[0]):
theta_smooth[k] = scipy.signal.savgol_filter(theta[k], smooth_win_len,
5)
theta_smooth[k] -= np.min(theta_smooth[k])
theta_smooth[k] /= np.max(theta_smooth[k])
plt.plot(theta_smooth.T)
plt.show()
theta_smooth *= -1
stim = utils.downsample(stim.squeeze(), downsample_ratio)
stim[stim > 0] = 1
stim_indices = np.where(np.diff(stim) == 1)[0] + 1
print('{} stim!'.format(len(stim_indices)))
plt.plot(stim)
plt.show()
data = np.vstack((stim, theta_smooth))
data = data[:, :, np.newaxis]
channel_list = [stim_channel] + channel_list
meta_data_fname = op.join(MMVT_DIR, subject, 'electrodes',
'electrodes_bipolar_data.npz')
np.savez(meta_data_fname,
data=data,
names=channel_list,
conditions=['on'],
dt=1 / fs)
def calc_cond_and_basline(subject,
con_method,
modality,
condition,
extract_mode,
band_name,
con_indentifer,
use_zvals,
node_names,
nodes_names_includes_hemi=False,
use_abs=True,
threshold=0.7,
window_length=25,
stc_downsample=2,
cond_name='interictals',
stc_subfolder='zvals',
stc_name=''):
import mne
from src.preproc import connectivity
input_fname, baseline_fname = get_cond_and_baseline_fnames(
subject, con_method, modality, condition, extract_mode, band_name,
con_indentifer, use_zvals, cond_name)
if not op.isfile(input_fname) or not op.isfile(baseline_fname):
# print('Can\'t find {}'.format(input_fname))
return None, None, None, None, None, None, None
stcs_fol = op.join(MMVT_DIR, subject, 'meg', stc_subfolder)
if stc_name == '':
stc_name = '{}-epilepsy-dSPM-meg-{}-average-amplitude-zvals-rh.stc'.format(
subject, condition)
stc_fname = op.join(stcs_fol, stc_name)
if op.isfile(stc_fname):
stc = mne.read_source_estimate(stc_fname)
times = utils.downsample(stc.times, stc_downsample) # [window_length:]
stc_data = np.max(stc.data, axis=0)
stc_data = utils.downsample(stc_data,
stc_downsample) # [window_length:]
else:
stc_data, times = None, None
d_cond, d_baseline = np.load(input_fname), np.load(baseline_fname)
con_values1, con_values2 = fix_con_values(d_cond)
con_values1, best_ords1 = connectivity.find_best_ord(con_values1,
return_ords=True)
con_values2, best_ords2 = connectivity.find_best_ord(con_values2,
return_ords=True)
# baseline_values1 = epi_utils.set_new_ords(d_baseline['con_values'], best_ords1)
# baseline_values2 = epi_utils.set_new_ords(d_baseline['con_values2'], best_ords2)
baseline_values1, baseline_values2 = fix_con_values(d_baseline)
baseline_values1 = connectivity.find_best_ord(baseline_values1,
return_ords=False)
baseline_values2 = connectivity.find_best_ord(baseline_values2,
return_ords=False)
mask1 = epi_utils.filter_connections(
con_values1,
d_cond['con_names'],
threshold,
node_names,
'',
use_abs,
nodes_names_includes_hemi=nodes_names_includes_hemi)
mask2 = epi_utils.filter_connections(
con_values2,
d_cond['con_names2'],
threshold,
node_names,
'',
use_abs,
nodes_names_includes_hemi=nodes_names_includes_hemi)
names = np.concatenate(
(d_cond['con_names'][mask1], d_cond['con_names2'][mask2]))
if len(names) == 0:
print('{} no connections'.format(condition))
return None, None, None, None, None, None, None
x_cond = np.concatenate((con_values1[mask1], con_values2[mask2]))
x_baseline = np.concatenate(
(baseline_values1[mask1], baseline_values2[mask2]))
if best_ords1 is not None and best_ords2 is not None:
best_ords = np.concatenate((best_ords1[mask1], best_ords2[mask2]))
names = [
'{} {}'.format(name, int(best_ord))
for name, best_ord in zip(names, best_ords)
]
return d_cond, d_baseline, x_cond, x_baseline, names, stc_data, times
def get_window_times(window_fname, downsample=2):
evoked = mne.read_evokeds(window_fname)[0]
times = evoked.times if len(evoked.times) % downsample == 0 else \
evoked.times[:-(downsample - 1)]
return utils.downsample(times, downsample)
def plot_modalities_power_spectrums_with_graph(subject,
modalities,
window_fname,
figure_name='',
percentiles=[5, 95],
inverse_method='dSPM',
ylims=[-18, 6],
file_type='jpg',
cb_ticks=[],
cb_ticks_font_size=12,
figure_fol=''):
evoked = mne.read_evokeds(window_fname)[0]
times = evoked.times if len(evoked.times) % 2 == 0 else evoked.times[:-1]
times = utils.downsample(times, 2)
min_t, max_t = round(times[0]), round(times[-1])
freqs = np.concatenate(
[np.arange(1, 30),
np.arange(31, 60, 3),
np.arange(60, 125, 5)])
# bands = dict(delta=[1, 4], theta=[4, 8], alpha=[8, 15], beta=[15, 30], gamma=[30, 55], high_gamma=[65, 120])
bands = dict(delta=[1, 4], high_gamma=[65, 120])
if figure_fol == '':
figure_fol = op.join(MMVT_DIR, subject, 'epilepsy-figures',
'power-spectrum')
fig, ax = plt.subplots(3, len(modalities), figsize=(20, 10))
for ind, modality in enumerate(modalities):
powers_ax = ax[0, ind]
powers_ax.set_title(modality.upper(), fontdict={'fontsize': 18})
positive_powers_ax = ax[1, ind]
negative_powers_ax = ax[2, ind]
root_dir = op.join(EEG_DIR if modality == 'eeg' else MEG_DIR, subject)
output_fname = op.join(
root_dir, '{}-epilepsy-{}-{}-{}-induced_norm_power.npz'.format(
subject, inverse_method, modality, '{window}'))
window = utils.namebase(window_fname)
window_output_fname = output_fname.format(window=window)
d = np.load(window_output_fname)
powers_negative, powers_positive = epi_utils.calc_masked_negative_and_positive_powers(
d['min'], d['max'], percentiles)
im1 = _plot_powers(powers_negative, powers_ax, times, freqs)
im2 = _plot_powers(powers_positive, powers_ax, times, freqs)
if ind == 0:
powers_ax.set_ylabel('Frequency (Hz)')
else:
powers_ax.set_yticks([])
add_colorbar(powers_ax, im2, cb_ticks, cb_ticks_font_size)
# for powers, axis, positive in zip([powers_positive, powers_negative], ax[1:2, ind], [True, False]):
# for band_name, band_freqs in bands.items():
# idx = [k for k, f in enumerate(freqs) if band_freqs[0] <= f <= band_freqs[1]]
# band_power = np.mean(powers[idx, :], axis=0)
# band_power[abs(band_power) < 2] = 0
# axis.plot(times, band_power.T, label=band_name)
# axis.set_xlim([min_t, max_t])
# axis.set_ylim([2, ylims[1]] if positive else [ylims[0], -2])
# if ind == 0:
# axis.set_ylabel('Positive Z-Scores' if positive else 'Negative Z-Scores')
# else:
# axis.set_yticks([])
# axis.legend(loc='center left', bbox_to_anchor=(1, 0.5))
# negative_powers_ax.set_xlabel('Time (s)')
for band_name, band_freqs in bands.items():
idx = [
k for k, f in enumerate(freqs)
if band_freqs[0] <= f <= band_freqs[1]
]
band_power = np.mean(powers_negative[idx, :], axis=0)
band_power[abs(band_power) < 2] = 0
positive_powers_ax.plot(times,
band_power.T,
label=band_name.replace('_', ' '))
positive_powers_ax.set_xlim([min_t, max_t])
positive_powers_ax.set_ylim([2, ylims[1]])
if ind == 0:
positive_powers_ax.set_ylabel('Positive Z-Scores')
else:
positive_powers_ax.set_yticks([])
if ind == len(modalities) - 1:
add_legend(positive_powers_ax)
for band_name, band_freqs in bands.items():
idx = [
k for k, f in enumerate(freqs)
if band_freqs[0] <= f <= band_freqs[1]
]
band_power = np.mean(powers_positive[idx, :], axis=0)
band_power[abs(band_power) < 2] = 0
negative_powers_ax.plot(times,
band_power.T,
label=band_name.replace('_', ' '))
negative_powers_ax.set_xlim([min_t, max_t])
negative_powers_ax.set_ylim([ylims[0], -2])
if ind == 0:
negative_powers_ax.set_ylabel('Negative Z-Scores')
else:
negative_powers_ax.set_yticks([])
if ind == len(modalities) - 1:
add_legend(negative_powers_ax)
negative_powers_ax.set_xlabel('Time (s)')
plt.tight_layout()
plt.subplots_adjust(left=None,
bottom=None,
right=0.92,
top=None,
wspace=None,
hspace=None)
if figure_name != '':
plt.savefig(op.join(figure_fol, '{}.{}'.format(figure_name,
file_type)),
dpi=300)
plt.close()
else:
plt.show()