Containers and data structure

NeurOne recordings consist of three structures:

  1. A recording containing (multiple)
  2. sessions containing (multiple)
  3. phases containing the actual data

Accordingly neurone_loader provides three containers representing these structures.

[2]:

from neurone_loader import Recording, Session, Phase

In each of these you can access the data and metadata like sampling_rate or channels and all of them support the features described in Lazy loading.

Session and Recording technically don’t hold data themselves. On the other hand one usually wants to work with the whole recording or at least a whole session of a recording. Therefore accessing the .data or any other attribute of Session or Recording will concatenate the data for you. Be aware that in order to save memory, accessing the .data attribute of a superseeding object will replace the .data attribute of the included containers with a view on the concatenated data. So if you manipulate any one of them you will also manipulate the other.

[3]:

rec = Recording(test_data_path)
session0 = rec.sessions[0]
sum_of_samples_of_phases_in_session0 = sum([p.n_samples for p in session0.phases])
print('\n')
print(f'session0.n_samples == sum_of_samples_of_phases_in_session0 is {session0.n_samples == sum_of_samples_of_phases_in_session0}')

(Lazy) loading Phase.n_samples
(Lazy) loading Phase.n_samples
(Lazy) loading Phase.n_samples
(Lazy) loading Phase.n_samples


session0.n_samples == sum_of_samples_of_phases_in_session0 is True

For a detailed description refer to the API documentation of the respective objects