def import_censor(self, ftup):
pdir = 'patient' + str(ftup[0]).zfill(3)
fname = (str(ftup[0]) + '.' + str(ftup[1]) + '.plx' + str(ftup[2]) +
'_censoring.mat')
fullname = self.datadir + pdir + '/' + fname
excludes = sio.loadmat(fullname)['excludes'].round(3)
if excludes.size != 0:
#do this in case some exclude ranges make no sense
badrng = np.where(np.diff(excludes, axis=1) < 0)
excludes = np.delete(excludes, badrng, axis=0)
# get data ready
ddict = {'patient': ftup[0], 'dataset': ftup[1],
'channel': ftup[2], 'start': excludes[:,0], 'stop': excludes[:,1]}
df = pd.DataFrame(ddict)
target = 'censor/' + make_path(*ftup)
self.write_to_db(target, df)
def import_lfp(self, ftup):
pdir = 'patient' + str(ftup[0]).zfill(3)
fname = str(ftup[0]) + '.' + str(ftup[1]) + '.plx' + str(ftup[2]) + '.mat'
fullname = self.datadir + pdir + '/' + fname
dset = h5py.File(fullname, 'r')
dat = dset['data'].value.squeeze()
sr = dset['srlfp'].value.squeeze()
vv = dat
desired_rate = 200. # desired final sampling rate (Hz)
decfrac = int(sr / desired_rate)
vv = decimate(dat, decfrac) # decimate data to 200 Hz
sr = sr / decfrac;
dt = (1. / sr).round(3)
times = (np.arange(0, vv.size) * dt).round(3).squeeze()
ddict = {'patient': ftup[0], 'dataset': ftup[1],
'channel': ftup[2], 'time': times, 'voltage': vv.values.squeeze()}
df = pd.DataFrame(ddict)
target = 'lfp/' + make_path(*ftup)
self.write_to_db(target, df)
def import_spikes(self, ftup):
pdir = 'patient' + str(ftup[0]).zfill(3)
fname = ('times_' + str(ftup[0]) + '.' + str(ftup[1]) + '.plx' +
str(ftup[2]) + '.mat')
fullname = self.datadir + pdir + '/' + fname
dat = sio.loadmat(fullname)['cluster_class']
unit = dat[:,0].astype('int')
times = np.around(dat[:,1]/1000., decimals=3) #times are originally in ms
sortord = np.argsort(times) #spikes aren't always time-sorted
times = times[sortord]
unit = unit[sortord]
# now restrict to valid units:
valid = unit == ftup[3]
times = times[valid]
unit = unit[valid]
ddict = {'patient': ftup[0], 'dataset': ftup[1],
'channel': ftup[2], 'unit': unit, 'time': times}
df = pd.DataFrame(ddict)
target = 'spikes/' + make_path(*ftup)
self.write_to_db(target, df)
def import_events(self, ftup):
trial_variables, events = self._grab_matlab_events(ftup)
evt = self._grab_plexon_events(ftup)
# get number of events
numtrials = max(events.index) + 1
# was this an FHC recording? if so, there are no Plexon stamps
# all events dumped into first or second slot, so other slots
# should have few timestamps
isFHC = (evt[2].size < 10)
if isFHC: # line up events with phys
# for now, we kludge this by just setting the clocks to be equal
# at task start and not worrying about drift
num_events = map(len, evt)
startcode = np.argmax(num_events) # these should be trial starts
# get time of first FHC event
FHC_start = evt[startcode][0].round(3).squeeze()
# compensate for offset
all_events = events.stack()
all_events.sort()
ephys_offset = (FHC_start - all_events.values[0]).round(3)
events['time'] += ephys_offset
else: # if we have Plexon events, use them
startcode = self.plx_codes['trial_start']
stopcode = self.plx_codes['trial_over']
# trial start -- sometimes a spurious event marks recording onset
if evt[startcode].shape[0] != numtrials:
evt[startcode] = evt[startcode][1:]
# trial stop -- when last trial aborted, may not be present
if evt[stopcode].shape[0] != numtrials:
evt[stopcode] = np.append(evt[stopcode], np.nan)
for var in self.plx_codes:
# valid = pd.notnull(events[var])
# events[var][valid] = evt[self.plx_codes[var]].round(3).squeeze()
this_selection = events['event'] == var
events['time'][this_selection] = evt[self.plx_codes[var]].round(3).squeeze()
# try to make events columns: this may fail in case some events
# can happen multiple times per trial; in that case, make each
# event a row and perform a join
if self.flatten_events:
# now merge task variables and events
df = events.join(trial_variables)
else:
# make event names column names
events = events.set_index('event', append=True).unstack()
# get rid of multi-index labeling
events.columns = pd.Index([e[1] for e in events.columns])
# now merge task variables and events
df = pd.concat([trial_variables, events], axis=1)
df['patient'] = ftup[0]
df['dataset'] = ftup[1]
# do some final tidying
df = df[df['result'] != 'aborted'] # get rid of aborts
target = 'events/' + make_path(*ftup[:-1])
self.write_to_db(target, df)