Help

Known issues

Problems in BIDS Manager {{ version }} that we know about, what you will see, and what to do until they are fixed. The pages of this site that meet one of them link here.

Found something not listed here? Report it.

In the converted dataset

The PET row of a scans table has a time without a date

What you see. Validation reports one error in the session's _scans.tsv: the PET row's acq_time is a time such as 09:28:23.000000, without a date. It happens with scanners whose PET header records an acquisition time but no acquisition date (the multimodal and PET samples' GE Advance).

What to do. Open the scans table in the Editor and type the date in front of the time, from the study date (2009-10-02T09:28:23 on the samples), or n/a. Validating in the Editor shows it step by step.

participants.tsv holds names and patient IDs

What you see. After a conversion from the GUI, participants.tsv has the columns given_name, family_name and patient_id, copied from the DICOM headers. The sidecars and spectroscopy headers have had these removed; the scan tables in .bidsmgr/, and a sourcedata/ copy if you keep one, still hold them.

What to do. Delete the three columns (and their entries in participants.json) before sharing, as in Sharing a dataset.

A dose sheet's none is read as empty

What you see. A PET dose sheet that gives none for ReconMethodParameterLabels and ReconMethodParameterUnits (what BIDS asks for when a reconstruction has no parameters) does not fill them: the import says it took fewer fields than the sheet has, and the sidecar gets a TODO placeholder, or nothing if placeholders are off.

What to do. After converting, set both fields to ["none"] in the PET sidecar (Editing metadata).

EEG and MEG names lose acq and rec, and run-01 becomes run-1

What you see. MNE-BIDS is given only the subject, session, task and run of a recording, so an acq- or rec- entity set in the table is not in the converted name, and a run is written without its leading zeros. The metadata you set for that recording may then not reach its sidecar. The BIDS preview marks such rows.

What to do. Avoid acq and rec on EEG and MEG rows for now, and add them after converting with Tools > Add or remove an entity...; Tools > Index widths... gives runs their leading zero back.

Files dcm2niix splits off keep dcm2niix's names

What you see. Some series are written as several files under dcm2niix's own endings (_e1, _e2, _ph, _real, _fieldmaphz, _ADC), which are not BIDS names; validation lists them as not part of the dataset. ASL series also miss fields no scanner records.

What to do. Rename or delete those files in the Editor, and fill the ASL fields in their sidecars.

A series the scan cannot name is skipped quietly

What you see. A series that no step could name stays ticked with an empty name; the table marks it as an error, but the conversion leaves it out without a line in the Log.

What to do. Check the Overview before converting: it lists them. Name them in the table, or untick them.

Smaller things in the output

  • An eye tracker's physiological events get an events.tsv stub from Generate missing companion files; delete it.
  • Scans tables list the .bval and .bvec files of a diffusion run as well as its image; delete those rows if a tool complains.
  • dataset_description.json > GeneratedBy > Container > Tag holds dcm2niix's usage line rather than a version.

In the Editor

An operation cannot be undone from the GUI

What you see. The Editor records every operation (a rename, a restructure, a deletion, a fix applied to many files, a saved file) with the files as they were, and several of its dialogs say the change can be undone. Undo in the toolbar, however, only steps back through the edits of the sidecar or table that is open. A defacing can be taken back with Tools > Put the face back...; other operations cannot be undone from the GUI.

What to do. Read each tool's preview before applying it. To restore a file by hand, find the operation in .bidsmgr/editor/operations.log and copy the file back from .bidsmgr/editor/originals/<operation>/ (The project folder and undo).

Check coherence offers to delete a MEG coordinate file

What you see. Check coherence lists the ..._coordsystem.json of a MEG or EEG recording as a sidecar with no recording, ticked for deletion. It is a required file of a MEG dataset.

What to do. Untick it before clicking Repair.

The Whole dataset view never shows its summary

The summary line of the Validation panel's Whole dataset view ("N findings in M kinds") is not shown. The list below it is complete.

In the table and the commands

bidsmgr-rebuild --from columns drops a DICOM row's task and run

What you see. After editing inventory.tsv in a spreadsheet and rebuilding from the columns, a BOLD run loses its task (an invalid name) and repeated series lose their run numbers. The scan keeps a DICOM row's task and run only in the entities column.

What to do. Rebuild from the entities, the default: bidsmgr-rebuild inventory.tsv. Edit the entities column, or curate in the GUI.

A hand-edited subject must be changed in two columns

What you see. In an inventory.tsv edited outside BIDS Manager, changing the subject in entities but not in participant_id (or the reverse) converts files named for one subject into the other subject's folder, without a warning. The GUI always changes both.

What to do. Change both columns together, or edit in the GUI.

An entity value that breaks the standard is only a warning

What you see. A value such as run-a, or a session with a space, is accepted in the table; the row is marked as a warning, and the name carries the bad value.

What to do. Read the issues of every row marked as a warning before converting.

In the GUI

The finished message counts failed subjects wrong

When subjects fail, the message at the end of a conversion says "1 subject(s) failed" whatever the number. The Log and .bidsmgr/errors/ have the full list.

The HTML validation report is in .bidsmgr/

Settings > Convert says the HTML report is written beside the dataset; it is written to .bidsmgr/validation_report.html inside it.

At a large font size in a small window

At font scale Comfortable (1.15x) or larger, in a window narrower than about 1600 pixels, the Converter's panels reach their minimum widths and Properties cuts its fields short; some message boxes cut their buttons' text. Fold the side panels you are not using, make the window wider, or use a smaller font scale.

The image viewer's shortcut list misses three keys

The list under the keyboard icon (or F1) of the image and spectrum viewers does not show Advanced controls (⌘I on macOS, Ctrl+I on Windows and Linux), Save a screenshot... (⇧⌘S on macOS, Ctrl+Shift+S on Windows and Linux) and Restore every viewer default..., which work. Viewers lists every key.

On some systems

Python 3.14: no defacing

niimath has no package for Python 3.14 yet. BIDS Manager installs and runs; Deface is greyed out with a tooltip saying why, and the quality check uses its own faster, less precise methods for the brain mask, the tissue classes and the registration. Use Python 3.10 to 3.13, or the installers, which use 3.10.

Reading EyeLink .edf files directly needs the Microsoft Visual C++ 2008 runtime. The Windows installer adds it; with pip, uv or conda, install it yourself (Install). Without it, an .asc export of the recording (SR Research's edf2asc) is read instead.