Converting MEG

The sample

Each recording is cropped to its first 25 seconds; everything else is as recorded.

MEG_Elekta_sample_data_small/
├── sub_us04rt22/
│   ├── 220209/   task_driving_run_01.fif, _run_02, task_rest.fif, task_emptypre.fif, task_emptypost.fif
│   └── 220215/   driving runs 1 to 6, rest, empty room before and after
└── sub_ye07us06/
    └── 220221/   driving runs 1 to 6, rest, empty room before and after

Scan

Step 1. Create a project and scan

Create a dataset named meg_study, choose the sample folder in Raw input, and click Scan…: 23 rows, in 2.0 seconds.

The MEG sample after the scan: 23 recordings of two participants, the first in two sessions found from the dates of its folders, each recording's task and run read from its file name. — The MEG sample after the scan.

What the scan read, and from where:

	From
Subject	the first folder: `sub_us04rt22` becomes `sub-001`, `sub_ye07us06` becomes `sub-002`
Session	the measurement dates in the files: the first participant was recorded on two days, so `ses-1` and `ses-2`; the second on one day, so no session
Task and run	the file names: `task_driving_run_05_00.fif` is task `driving`, run 5 (the trailing `_00` is the recording system's numbering, dropped)

Step 2. Read the Overview

Overview points out that sub-002 has no session while sub-001 has two. If the two days of sub-001 are two conditions of your study, keep them; if every participant came once and sub-001 came back for a reason of its own, consider giving each participant one session per day consistently.

Empty-room recordings

The empty-room recordings measure the noise of the room and the sensors with nobody in the helmet, before and after each session. BIDS names them with the task noise; here they arrived as the tasks emptypre and emptypost.

Step 3. Rename them

Select the emptypre and emptypost rows, set task to noise with Bulk edit…, then set run to 1 for the ones recorded before and 2 for the ones after.

An empty-room recording after renaming: task noise, run 1 (before the session's recordings; run 2 is the one after). — An empty-room recording after renaming.

They stay in each participant's session folder, beside the recordings of that day. BIDS also allows a subject of their own, sub-emptyroom, with one session per date; either way, a recording names its empty room in its sidecar's AssociatedEmptyRoom field, which you can fill in the Editor after converting.

Describe, convert and check

Step 4. Describe and convert

In Dataset metadata…, give the authors and the licence. The MEG section of the template asks only for the MEG sidecar fields the conversion cannot read from the file, such as DewarPosition, AssociatedEmptyRoom and SubjectArtefactDescription. The manufacturer, the line frequency (50 Hz here) and the sensor geometry come from the FIF file itself. Click Run conversion: the sample converted in 5.3 seconds into 81 files.

Step 5. Validate

Validate dataset in the Editor reports 33 files valid, 361 warnings and 0 errors; the warnings are mostly the 277 TODO placeholders.

Step 6. Look at a recording

Select a driving run and click Load signal. Choose the magnetometers in the channel types, and normalise the channels to see each one's shape.

The first driving run of the second session, converted: 24 of the magnetometers, filtered from 1 to 40 Hz, each scaled to itself. — A driving run of the second session, converted.

QC checks the recording channel by channel; a noisy or flat channel can be marked bad and saved into channels.tsv (Checking image quality).

In your own MEG data

Session from dates: a participant recorded on two days gets two sessions. A folder named ses-<label> overrides it.
CTF recordings (.ds folders) and KIT (.con, .sqd) are read the same way.
Empty-room recordings are easiest to use when their name says noise, and each recording names its own in AssociatedEmptyRoom.