Chapter 5: Sensor-Level Analysis¶
Continuing from Chapter 4 — Preprocessing.
"Sensor-level" means working directly with what the electrodes or sensors measured, before any attempt to localize it inside the brain — the same stage MNE-Python's Events and Epochs and ERP Analysis chapters cover.
Events and epoching¶
Brainstorm detects trigger events the same way MNE-Python does — by reading a stimulus channel or an accompanying events file. Once events are detected and listed in the tree, a process cuts the continuous recording into short epochs around each event of interest: one snippet per trial, the direct equivalent of MNE-Python's Epochs object.
Averaging into an ERP/ERF¶
Running the "Average" process over a set of epochs produces one averaged waveform — an ERP for EEG or ERF for MEG — exactly what MNE-Python calls an Evoked object. Double-clicking the resulting file opens it as a stacked time series (one line per channel) or, with one more click, as a scalp topography: a color map showing the signal's spatial pattern across the head at a chosen instant.
Comparing conditions¶
Because every result is just a file in the tree, comparing two experimental conditions is usually: average each condition separately, then either view both side by side in the Viewer, or run a "Difference" process to compute condition A minus condition B directly as a new file.
A note on time-frequency¶
Brainstorm also offers sensor-level time-frequency analysis (Morlet wavelets and related methods) as a process, producing a time × frequency map per channel — useful for rhythmic activity that a simple average would cancel out. It isn't covered in depth in this beginner track, but it lives in the same process list as everything above.