def test_filter_events_array():
with pytest.raises(ValueError):
filter_events_array(arr=np.array([0.0, 0.0, 0.6]))
with pytest.raises(ValueError):
filter_events_array(arr=np.array([[0.0, 0.0, 0.6]]), n_time_steps=0)
arr = np.array([[0.0, 0.0, 0.6]])
filtered_events_array = filter_events_array(arr=arr)
assert arr.shape[0] == filtered_events_array.shape[0]
assert arr.shape[1] == filtered_events_array.shape[1]
expected = np.array([[0.0, 0.0, 0.199559]])
assert (np.abs(filtered_events_array - expected) < 1e-6).all()
def __init__(self,
events: np.ndarray,
events_meta: pd.DataFrame,
filter_scale: float = 2,
filter_n_time_steps: int = 20):
"""
Parameters
----------
events
events
events_meta
lambda, noise_std, cell_roi_id for each roi
filter_scale
See filter_events_array for description
filter_n_time_steps
See filter_events_array for description
"""
filtered_events = filter_events_array(arr=events,
scale=filter_scale,
n_time_steps=filter_n_time_steps)
# Convert matrix to list of 1d arrays so that it can be stored
# in a single column of the dataframe
events = [x for x in events]
filtered_events = [x for x in filtered_events]
df = pd.DataFrame({
'events': events,
'filtered_events': filtered_events,
'lambda': events_meta['lambda'],
'noise_std': events_meta['noise_std'],
'cell_roi_id': events_meta['cell_roi_id']
})
super().__init__(name='events', value=df)
def get_events(self,
filter_scale: float = 2,
filter_n_time_steps: int = 20) -> pd.DataFrame:
"""
Parameters
----------
filter_scale: float
See filter_events_array for description
filter_n_time_steps: int
See filter_events_array for description
Returns
-------
Events dataframe:
columns:
events: np.array
lambda: float
noise_std: float
cell_roi_id: int
index:
cell_specimen_id: int
"""
event_detection = self.nwbfile.processing['ophys']['event_detection']
# NOTE: The rois with events are stored in event detection
partial_cell_specimen_table = event_detection.rois.to_dataframe()
events = event_detection.data[:]
# events stored time x roi. Change back to roi x time
events = events.T
filtered_events = filter_events_array(arr=events,
scale=filter_scale,
n_time_steps=filter_n_time_steps)
# Convert to list to that it can be stored in a single column
events = [x for x in events]
filtered_events = [x for x in filtered_events]
return pd.DataFrame(
{
'cell_roi_id': partial_cell_specimen_table.index,
'events': events,
'filtered_events': filtered_events,
'lambda': event_detection.lambdas[:],
'noise_std': event_detection.noise_stds[:]
},
index=pd.Index(partial_cell_specimen_table['cell_specimen_id']))
def get_events(self,
filter_scale: float = 2,
filter_n_time_steps: int = 20) -> pd.DataFrame:
"""
Returns events in dataframe format
Parameters
----------
filter_scale: float
See filter_events_array for description
filter_n_time_steps: int
See filter_events_array for description
See behavior_ophys_session.events for return type
"""
events_file = self.extractor.get_event_detection_filepath()
with h5py.File(events_file, 'r') as f:
events = f['events'][:]
lambdas = f['lambdas'][:]
noise_stds = f['noise_stds'][:]
roi_ids = f['roi_names'][:]
filtered_events = filter_events_array(arr=events,
scale=filter_scale,
n_time_steps=filter_n_time_steps)
# Convert matrix to list of 1d arrays so that it can be stored
# in a single column of the dataframe
events = [x for x in events]
filtered_events = [x for x in filtered_events]
df = pd.DataFrame({
'events': events,
'filtered_events': filtered_events,
'lambda': lambdas,
'noise_std': noise_stds,
'cell_roi_id': roi_ids
})
# Set index as cell_specimen_id from cell_specimen_table
cell_specimen_table = self.get_cell_specimen_table()
cell_specimen_table = cell_specimen_table.reset_index()
df = cell_specimen_table[['cell_roi_id', 'cell_specimen_id']]\
.merge(df, on='cell_roi_id')
df = df.set_index('cell_specimen_id')
return df