def _initialize(self, **kargs):
QOscilloscope._initialize(self, **kargs)
self.inputs['spikes'].set_buffer(size=self.peak_buffer_size,
double=False)
# poller onpeak
self._last_peak = 0
self.poller_peak = ThreadPollInput(input_stream=self.inputs['spikes'],
return_data=True)
self.poller_peak.new_data.connect(self._on_new_peak)
self.spikes_array = self.inputs['spikes'].buffer.buffer
self.scatters = {}
self.change_catalogue(self.catalogue)
self.params['xsize'] = 1.
self.params['decimation_method'] = 'min_max'
self.params['mode'] = 'scan'
self.params['scale_mode'] = 'same_for_all'
self.params['display_labels'] = True
self.timer_scale = QT.QTimer(singleShot=True, interval=500)
self.timer_scale.timeout.connect(self.auto_scale)
self.timer_scale.start()
def __init__(self, input_stream, output_streams, peeler,in_group_channels,
timeout = 200, parent = None):
ThreadPollInput.__init__(self, input_stream, timeout=timeout, return_data=True, parent = parent)
self.output_streams = output_streams
self.peeler = peeler
self.in_group_channels = in_group_channels
self.sample_rate = input_stream.params['sample_rate']
self.total_channel = self.input_stream().params['shape'][1]
def __init__(self, input_stream, output_streams, peeler,in_group_channels,
timeout = 200, parent = None):
ThreadPollInput.__init__(self, input_stream, timeout=timeout, return_data=True, parent = parent)
self.output_streams = output_streams
self.peeler = peeler
self.in_group_channels = in_group_channels
self.sample_rate = input_stream.params['sample_rate']
self.total_channel = self.input_stream().params['shape'][1]
self.mutex = Mutex()
def _initialize(self, **kargs):
self.sample_rate = self.inputs['signals'].params['sample_rate']
self.wf_dtype = self.inputs['signals'].params['dtype']
self.inputs['spikes'].set_buffer(size=self.peak_buffer_size,
double=False)
buffer_sigs_size = int(self.sample_rate * 3.)
self.inputs['signals'].set_buffer(size=buffer_sigs_size, double=False)
# poller
self.poller_sigs = ThreadPollInput(input_stream=self.inputs['signals'],
return_data=True)
self.poller_spikes = ThreadPollInput(
input_stream=self.inputs['spikes'], return_data=True)
self.histogram_2d = {}
self.last_waveform = {}
self.change_catalogue(self.catalogue)
self.timer = QT.QTimer(interval=100)
self.timer.timeout.connect(self.refresh)
def _initialize(self, **kargs):
QOscilloscope._initialize(self, **kargs)
self.inputs['spikes'].set_buffer(size=self.peak_buffer_size, double=False)
# poller onpeak
self._last_peak = 0
self.poller_peak = ThreadPollInput(input_stream=self.inputs['spikes'], return_data=True)
self.poller_peak.new_data.connect(self._on_new_peak)
self.spikes_array = self.inputs['spikes'].buffer.buffer
self._default_color = QtGui.QColor('#FFFFFF')#TODO
self.scatters = {}
for k in self.catalogue['cluster_labels']:
color = self.catalogue['cluster_colors'].get(k, (1,1,1))
r, g, b = color
qcolor = QtGui.QColor(r*255, g*255, b*255)
qcolor.setAlpha(150)
scatter = pg.ScatterPlotItem(x=[ ], y= [ ], pen=None, brush=qcolor, size=10, pxMode = True)
self.scatters[k] = scatter
self.plot.addItem(scatter)
class OnlineWaveformHistViewer(WidgetNode):
_input_specs = {
'signals': dict(streamtype='signals'),
'spikes': dict(streamtype='events', shape=(-1, ), dtype=_dtype_spike),
}
_params = [
{
'name': 'colormap',
'type': 'list',
'values': [
'hot',
'viridis',
'jet',
'gray',
]
},
{
'name': 'bin_min',
'type': 'float',
'value': -20.
},
{
'name': 'bin_max',
'type': 'float',
'value': 8.
},
{
'name': 'bin_size',
'type': 'float',
'value': .1
},
{
'name': 'refresh_interval',
'type': 'int',
'value': 100,
'limits': [5, 1000]
},
]
def __init__(self, **kargs):
WidgetNode.__init__(self, **kargs)
self.layout = QT.QVBoxLayout()
self.setLayout(self.layout)
h = QT.QHBoxLayout()
self.layout.addLayout(h)
self.combobox = QT.QComboBox()
h.addWidget(self.combobox)
but = QT.QPushButton('clear')
h.addWidget(but)
but.clicked.connect(self.on_clear)
self.label = QT.QLabel('')
h.addWidget(self.label)
self.graphicsview = pg.GraphicsView()
self.layout.addWidget(self.graphicsview)
self.params = pg.parametertree.Parameter.create(name='settings',
type='group',
children=self._params)
self.tree_params = pg.parametertree.ParameterTree(parent=self)
self.tree_params.header().hide()
self.tree_params.setParameters(self.params, showTop=True)
self.tree_params.setWindowTitle('Options for waveforms hist viewer')
self.tree_params.setWindowFlags(QT.Qt.Window)
self.params.sigTreeStateChanged.connect(self.on_params_changed)
self.initialize_plot()
self.mutex = Mutex()
def _configure(self, peak_buffer_size=100000, catalogue=None, **kargs):
self.peak_buffer_size = peak_buffer_size
self.catalogue = catalogue
def _initialize(self, **kargs):
self.sample_rate = self.inputs['signals'].params['sample_rate']
self.wf_dtype = self.inputs['signals'].params['dtype']
self.inputs['spikes'].set_buffer(size=self.peak_buffer_size,
double=False)
buffer_sigs_size = int(self.sample_rate * 3.)
self.inputs['signals'].set_buffer(size=buffer_sigs_size, double=False)
# poller
self.poller_sigs = ThreadPollInput(input_stream=self.inputs['signals'],
return_data=True)
self.poller_spikes = ThreadPollInput(
input_stream=self.inputs['spikes'], return_data=True)
self.histogram_2d = {}
self.last_waveform = {}
self.change_catalogue(self.catalogue)
self.timer = QT.QTimer(interval=100)
self.timer.timeout.connect(self.refresh)
def _start(self, **kargs):
self.last_head_sigs = None
self.last_head_spikes = None
self.timer.start()
self.inputs['signals'].empty_queue()
self.inputs['spikes'].empty_queue()
self.poller_sigs.start()
self.poller_spikes.start()
def _stop(self, **kargs):
self.timer.stop()
self.poller_sigs.stop()
self.poller_sigs.wait()
self.poller_spikes.stop()
self.poller_spikes.wait()
def _close(self, **kargs):
pass
def open_settings(self):
if not self.tree_params.isVisible():
self.tree_params.show()
else:
self.tree_params.hide()
def on_params_changed(self, params, changes):
self.change_lut()
self.change_catalogue(self.catalogue)
self.timer.setInterval(self.params['refresh_interval'])
def initialize_plot(self):
self.viewBox = MyViewBox()
self.viewBox.doubleclicked.connect(self.open_settings)
self.viewBox.gain_zoom.connect(self.gain_zoom)
self.viewBox.disableAutoRange()
self.plot = pg.PlotItem(viewBox=self.viewBox)
self.graphicsview.setCentralItem(self.plot)
self.plot.hideButtons()
self.image = pg.ImageItem()
self.plot.addItem(self.image)
self.curve_spike = pg.PlotCurveItem()
self.plot.addItem(self.curve_spike)
self.curve_limit = pg.PlotCurveItem()
self.plot.addItem(self.curve_limit)
self.change_lut()
def change_lut(self):
N = 512
cmap_name = self.params['colormap']
cmap = matplotlib.cm.get_cmap(cmap_name, N)
lut = []
for i in range(N):
r, g, b, _ = matplotlib.colors.ColorConverter().to_rgba(cmap(i))
lut.append([r * 255, g * 255, b * 255])
self.lut = np.array(lut, dtype='uint8')
def change_catalogue(self, catalogue):
self.params.blockSignals(True)
with self.mutex:
self.catalogue = catalogue
colors = make_color_dict(self.catalogue['clusters'])
self.qcolors = {}
for k, color in colors.items():
r, g, b = color
self.qcolors[k] = QT.QColor(r * 255, g * 255, b * 255)
self.all_plotted_labels = self.catalogue['cluster_labels'].tolist(
) + [LABEL_UNCLASSIFIED]
centers0 = self.catalogue['centers0']
if centers0.shape[0] > 0:
self.params['bin_min'] = np.min(centers0) * 1.5
self.params['bin_max'] = np.max(centers0) * 1.5
bin_min, bin_max = self.params['bin_min'], self.params['bin_max']
bin_size = self.params['bin_size']
self.bins = np.arange(bin_min, bin_max, self.params['bin_size'])
self.combobox.clear()
self.combobox.addItems([str(k) for k in self.all_plotted_labels])
self.on_clear()
self._max = 10
_, peak_width, nb_chan = self.catalogue['centers0'].shape
x, y = [], []
for c in range(1, nb_chan):
x.extend([c * peak_width, c * peak_width, np.nan])
y.extend([-1000, 1000, np.nan])
self.curve_limit.setData(x=x, y=y, connect='finite')
self.params.blockSignals(False)
def on_clear(self):
with self.mutex:
shape = self.catalogue['centers0'].shape
self.indexes0 = np.arange(shape[1] * shape[2], dtype='int64')
self.histogram_2d = {}
self.last_waveform = {}
self.nb_spikes = {}
for k in self.all_plotted_labels:
self.histogram_2d[k] = np.zeros(
(shape[1] * shape[2], self.bins.size), dtype='int64')
self.last_waveform[k] = np.zeros((shape[1] * shape[2], ),
dtype=self.wf_dtype)
self.nb_spikes[k] = 0
self.plot.setXRange(0, self.indexes0[-1] + 1)
self.plot.setYRange(self.params['bin_min'], self.params['bin_max'])
def auto_scale(self):
pass
def gain_zoom(self, v):
self._max *= v
self.image.setLevels([0, self._max], update=True)
def refresh(self):
#~ print('refresh')
#~ t0 = time.perf_counter()
head_sigs = self.poller_sigs.pos()
head_spikes = self.poller_spikes.pos()
if self.last_head_sigs is None:
self.last_head_sigs = head_sigs
if self.last_head_spikes is None:
self.last_head_spikes = head_spikes
if self.last_head_spikes is None or self.last_head_sigs is None:
return
# update image
n_right, n_left = self.catalogue['n_right'], self.catalogue['n_left']
bin_min, bin_max, bin_size = self.params['bin_min'], self.params[
'bin_max'], self.params['bin_size']
# check peak_buffer_size here
if (head_spikes - self.last_head_spikes) >= (0.9 *
self.peak_buffer_size):
self.last_head_spikes = head_spikes - int(
0.9 * self.peak_buffer_size)
new_spikes = self.inputs['spikes'].get_data(self.last_head_spikes,
head_spikes)
right_indexes = new_spikes['index'] + n_right
if np.any(right_indexes > head_sigs):
# the buffer of signals is available for some spikes yet
# so remove then for this loop and get back on head_spikes
first_out = np.nonzero(right_indexes)[0][0]
head_spikes = head_spikes - (new_spikes.size - first_out)
new_spikes = new_spikes[:first_out]
for k in self.all_plotted_labels:
mask = new_spikes['cluster_label'] == k
indexes = new_spikes[mask]['index']
for ind in indexes:
wf = self.inputs['signals'].get_data(ind + n_left,
ind + n_right)
wf = wf.T.reshape(-1)
wf_bined = np.floor((wf - bin_min) / bin_size).astype('int32')
wf_bined = wf_bined.clip(0, self.bins.size - 1)
with self.mutex:
self.histogram_2d[k][self.indexes0, wf_bined] += 1
self.last_waveform[k] = wf
self.nb_spikes[k] += 1
self.last_head_sigs = head_sigs
self.last_head_spikes = head_spikes
if self.combobox.currentIndex() == -1:
return
if self.visibleRegion().isEmpty():
# when several tabs not need to refresh
return
# refresh plot , update image
k = self.all_plotted_labels[self.combobox.currentIndex()]
hist2d = self.histogram_2d[k]
self.image.setImage(hist2d, lut=self.lut, levels=[0, self._max])
self.image.setRect(
QT.QRectF(-0.5, bin_min, hist2d.shape[0], bin_max - bin_min))
self.image.show()
self.curve_spike.setData(x=self.indexes0,
y=self.last_waveform[k],
pen=pg.mkPen(self.qcolors[k], width=1.5))
txt = 'nbs_pike = {}'.format(self.nb_spikes[k])
self.label.setText(txt)
class OnlineTraceViewer(QOscilloscope):
_input_specs = {'signals': dict(streamtype='signals'),
'spikes': dict(streamtype='events', shape = (-1, ), dtype=_dtype_spike),
}
_default_params = QOscilloscope._default_params
def __init__(self, **kargs):
QOscilloscope.__init__(self, **kargs)
def _configure(self, peak_buffer_size = 10000, catalogue=None, **kargs):
QOscilloscope._configure(self, **kargs)
self.peak_buffer_size = peak_buffer_size
self.catalogue = catalogue
assert catalogue is not None
def _initialize(self, **kargs):
QOscilloscope._initialize(self, **kargs)
self.inputs['spikes'].set_buffer(size=self.peak_buffer_size, double=False)
# poller onpeak
self._last_peak = 0
self.poller_peak = ThreadPollInput(input_stream=self.inputs['spikes'], return_data=True)
self.poller_peak.new_data.connect(self._on_new_peak)
self.spikes_array = self.inputs['spikes'].buffer.buffer
self._default_color = QtGui.QColor('#FFFFFF')#TODO
self.scatters = {}
for k in self.catalogue['cluster_labels']:
color = self.catalogue['cluster_colors'].get(k, (1,1,1))
r, g, b = color
qcolor = QtGui.QColor(r*255, g*255, b*255)
qcolor.setAlpha(150)
scatter = pg.ScatterPlotItem(x=[ ], y= [ ], pen=None, brush=qcolor, size=10, pxMode = True)
self.scatters[k] = scatter
self.plot.addItem(scatter)
#~ for i in range(self.nb_channel):
#~ color = self._default_color
#~ color.setAlpha(150)
#~ scatter = pg.ScatterPlotItem(x=[ ], y= [ ], pen=None, brush=color, size=10, pxMode = True)
#~ self.scatters.append(scatter)
#~ self.plot.addItem(scatter)
def _start(self, **kargs):
QOscilloscope._start(self, **kargs)
self._last_peak = 0
self.poller_peak.start()
def _stop(self, **kargs):
QOscilloscope._stop(self, **kargs)
self.poller_peak.stop()
self.poller_peak.wait()
def _close(self, **kargs):
QOscilloscope._close(self, **kargs)
def reset_curves_data(self):
QOscilloscope.reset_curves_data(self)
self.t_vect_full = np.arange(0,self.full_size, dtype=float)/self.sample_rate
self.t_vect_full -= self.t_vect_full[-1]
def _on_new_peak(self, pos, data):
self._last_peak = pos
def autoestimate_scales(self):
# in our case preprocesssed signal is supposed to be normalized
self.all_mean = np.zeros(self.nb_channel,)
self.all_sd = np.ones(self.nb_channel,)
return self.all_mean, self.all_sd
def _refresh(self, **kargs):
QOscilloscope._refresh(self, **kargs)
mode = self.params['mode']
gains = np.array([p['gain'] for p in self.by_channel_params.children()])
offsets = np.array([p['offset'] for p in self.by_channel_params.children()])
visibles = np.array([p['visible'] for p in self.by_channel_params.children()], dtype=bool)
head = self._head
full_arr = self.inputs['signals'].get_data(head-self.full_size, head)
if self._last_peak==0:
return
keep = (self.spikes_array['index']>head - self.full_size) & (self.spikes_array['index']<head)
spikes = self.spikes_array[keep]
#~ spikes = self.spikes_array['index'][keep]
spikes_ind = spikes['index'] - (head - self.full_size)
spikes_amplitude = full_arr[spikes_ind, :]
spikes_amplitude[:, visibles] *= gains[visibles]
spikes_amplitude[:, visibles] += offsets[visibles]
if mode=='scroll':
peak_times = self.t_vect_full[spikes_ind]
elif mode =='scan':
#some trick to play with fake time
front = head % self.full_size
ind1 = (spikes['index']%self.full_size)<front
ind2 = (spikes['index']%self.full_size)>front
peak_times = self.t_vect_full[spikes_ind]
peak_times[ind1] += (self.t_vect_full[front] - self.t_vect_full[-1])
peak_times[ind2] += (self.t_vect_full[front] - self.t_vect_full[0])
for i, k in enumerate(self.catalogue['cluster_labels']):
keep = k==spikes['label']
if np.sum(keep)>0:
chan = self.catalogue['max_on_channel'][i]
if visibles[chan]:
self.scatters[k].setData(peak_times[keep], spikes_amplitude[keep, chan])
else:
self.scatters[k].setData([], [])
else:
self.scatters[k].setData([], [])
class OnlineTraceViewer(QOscilloscope):
_input_specs = {
'signals': dict(streamtype='signals'),
'spikes': dict(streamtype='events', shape=(-1, ), dtype=_dtype_spike),
}
_default_params = QOscilloscope._default_params
def __init__(self, **kargs):
QOscilloscope.__init__(self, **kargs)
self.mutex = Mutex()
def _configure(self, peak_buffer_size=100000, catalogue=None, **kargs):
QOscilloscope._configure(self, **kargs)
self.peak_buffer_size = peak_buffer_size
self.catalogue = catalogue
assert catalogue is not None
def _initialize(self, **kargs):
QOscilloscope._initialize(self, **kargs)
self.inputs['spikes'].set_buffer(size=self.peak_buffer_size,
double=False)
# poller onpeak
self._last_peak = 0
self.poller_peak = ThreadPollInput(input_stream=self.inputs['spikes'],
return_data=True)
self.poller_peak.new_data.connect(self._on_new_peak)
self.spikes_array = self.inputs['spikes'].buffer.buffer
self.scatters = {}
self.change_catalogue(self.catalogue)
self.params['xsize'] = 1.
self.params['decimation_method'] = 'min_max'
self.params['mode'] = 'scan'
self.params['scale_mode'] = 'same_for_all'
self.params['display_labels'] = True
self.timer_scale = QT.QTimer(singleShot=True, interval=500)
self.timer_scale.timeout.connect(self.auto_scale)
self.timer_scale.start()
def _start(self, **kargs):
QOscilloscope._start(self, **kargs)
self._last_peak = 0
self.poller_peak.start()
def _stop(self, **kargs):
QOscilloscope._stop(self, **kargs)
self.poller_peak.stop()
self.poller_peak.wait()
def _close(self, **kargs):
QOscilloscope._close(self, **kargs)
def reset_curves_data(self):
QOscilloscope.reset_curves_data(self)
self.t_vect_full = np.arange(0, self.full_size,
dtype=float) / self.sample_rate
self.t_vect_full -= self.t_vect_full[-1]
def _on_new_peak(self, pos, data):
self._last_peak = pos
def autoestimate_scales(self):
# in our case preprocesssed signal is supposed to be normalized
self.all_mean = np.zeros(self.nb_channel, )
self.all_sd = np.ones(self.nb_channel, )
return self.all_mean, self.all_sd
def change_catalogue(self, catalogue):
with self.mutex:
for k, v in self.scatters.items():
self.plot.removeItem(v)
self.scatters = {}
self.catalogue = catalogue
colors = make_color_dict(self.catalogue['clusters'])
self.qcolors = {}
for k, color in colors.items():
r, g, b = color
self.qcolors[k] = QT.QColor(r * 255, g * 255, b * 255)
self.all_plotted_labels = self.catalogue['cluster_labels'].tolist(
) + [LABEL_UNCLASSIFIED]
for k in self.all_plotted_labels:
qcolor = self.qcolors[k]
qcolor.setAlpha(150)
scatter = pg.ScatterPlotItem(x=[],
y=[],
pen=None,
brush=qcolor,
size=10,
pxMode=True)
self.scatters[k] = scatter
self.plot.addItem(scatter)
def _refresh(self, **kargs):
if self.visibleRegion().isEmpty():
# when several tabs not need to refresh
return
with self.mutex:
QOscilloscope._refresh(self, **kargs)
mode = self.params['mode']
gains = np.array(
[p['gain'] for p in self.by_channel_params.children()])
offsets = np.array(
[p['offset'] for p in self.by_channel_params.children()])
visibles = np.array(
[p['visible'] for p in self.by_channel_params.children()],
dtype=bool)
head = self._head
full_arr = self.inputs['signals'].get_data(head - self.full_size,
head)
if self._last_peak == 0:
return
keep = (self.spikes_array['index'] > head - self.full_size) & (
self.spikes_array['index'] < head)
spikes = self.spikes_array[keep]
spikes_ind = spikes['index'] - (head - self.full_size)
spikes_ind = spikes_ind[spikes_ind < full_arr.shape[
0]] # to avoid bug if last peak is great than head
real_spikes_amplitude = full_arr[spikes_ind, :]
spikes_amplitude = real_spikes_amplitude.copy()
spikes_amplitude[:, visibles] *= gains[visibles]
spikes_amplitude[:, visibles] += offsets[visibles]
if mode == 'scroll':
peak_times = self.t_vect_full[spikes_ind]
elif mode == 'scan':
#some trick to play with fake time
front = head % self.full_size
ind1 = (spikes['index'] % self.full_size) < front
ind2 = (spikes['index'] % self.full_size) > front
peak_times = self.t_vect_full[spikes_ind]
peak_times[ind1] += (self.t_vect_full[front] -
self.t_vect_full[-1])
peak_times[ind2] += (self.t_vect_full[front] -
self.t_vect_full[0])
for i, k in enumerate(self.all_plotted_labels):
keep = k == spikes['cluster_label']
if np.sum(keep) > 0:
if k >= 0:
chan = self.catalogue['extremum_channel'][i]
if visibles[chan]:
times, amps = peak_times[keep], spikes_amplitude[
keep, chan]
else:
times, amps = [], []
else:
chan_max = np.argmax(np.abs(
real_spikes_amplitude[keep, :]),
axis=1)
keep2 = visibles[chan_max]
chan_max = chan_max[keep2]
keep[keep] &= keep2
times, amps = peak_times[keep], spikes_amplitude[
keep, chan_max]
self.scatters[k].setData(times, amps)
else:
self.scatters[k].setData([], [])
def auto_scale(self, spacing_factor=25.):
self.params_controller.compute_rescale(spacing_factor=spacing_factor)
self.refresh()
