BIDS Manager is a desktop application that turns raw MRI, PET, EEG, MEG and iEEG recordings, together with PET blood curves and the physiological traces Siemens CMRR sequences write beside an MRI series, into a validated BIDS dataset.
It scans your raw data, shows you every conversion decision in a table you can edit, runs the conversion, and then opens the result for metadata editing, signal and volume inspection, and validation. One application, no scripts, no hand-editing JSON.
The thing it is really built around is not the conversion. Converting a DICOM
series is a solved problem, and BIDS Manager uses the same engines everyone else
does: dcm2niix for MRI and PET DICOM, mne-bids for electrophysiology,
nibabel for ECAT, bidsphysio for physio, and pet2bids for blood curves and
the PET metadata no converter writes. What it adds is everything around them:
seeing what you have before anything is written, saying what the files cannot
say for themselves, and being told what is wrong in terms of the standard rather
than a stack trace.
The quickest route is the one-click bootstrap installer for macOS, Linux and Windows. It bundles a portable Python with every dependency and registers a native desktop launcher, so no existing Python install is required. The install guide walks through it.
With Python already set up, pip install bids-manager works too.
Launch the interface with bidsmgr. Prefer the command line? Seven verbs cover
the whole pipeline:
bidsmgr-create scaffold a dataset and its project
bidsmgr-scan walk a raw tree and build the inventory
bidsmgr-rebuild rebuild BIDS names from edited entities
bidsmgr-convert convert, routing each row to the right engine
bidsmgr-metadata dataset_description, participants, phenotype
bidsmgr-validate validate, with a report you can hand to a colleague
bidsmgr-project list a project's saved scan versions
The documentation has the full GUI walkthrough, a reference for every flag, and a tutorial per modality with a sample dataset you can download and work through.
The features below follow those steps.
Point the application at a folder of DICOM, EEG, MEG or PET recordings, or all of them at once. It walks the tree, works out what each series is, and shows the proposed BIDS name for every one in a table you can sort, filter and bulk-edit.
Formats are recognised by reading the file, not by its extension, so a Philips export whose filename is a bare identifier still converts and a renamed ECAT is still an ECAT. Anything it sets aside says why: a scanner report with no image data in it, a localiser, a series it cannot classify.
Change any cell before you commit. Subjects, sessions, tasks, runs: the BIDS filename updates as you type, so what you see is what will be written.
Two recordings that would land on the same filename are caught before anything
is written. Genuine repeats are given a run number where the standard allows
one; where it does not, both are shown in red and the conversion refuses to
start rather than write one file over another.
Conversion runs per subject into a staging folder and is committed only when that subject finishes, so a failure never leaves a half-converted tree.
Some things are simply not in the data. An EEG file has nowhere to record its reference or its ground. A PET scanner records how it reconstructed an image but not how much tracer went into the person, in what form, or when.
The metadata form is generated from the BIDS schema, so it asks exactly what the standard declares for each kind of file, at its real requirement level, with the standard's own description on hover. Answer once for the study, override for the one recording that differs. What the conversion already worked out is folded away, so you are only asked what nobody could answer for you.
Every sidecar opens as a schema-aware form and every table as a spreadsheet, so correcting a converted dataset does not mean editing JSON by hand. Edits are undoable, and validation can be re-run against them without leaving the window.
Viewers come with it. Images open one plane at a time, as three planes sharing a crosshair, or in a GPU renderer with clipping, lighting and colour-FA. A 4-D run gains a time-series graph, and on PET its axis is real seconds taken from the frame times, because PET frames are not evenly spaced and a frame index flattens the part worth looking at.
EEG, MEG and iEEG open as an interactive signal viewer with channel filtering,
per-segment filtering, an in-application power spectrum and events overlaid from
the events.tsv beside them.
Validation is part of the same application, reads the same BIDS schema the metadata form was built from, and runs on a dataset of any modality at once.
Every finding names the schema rule it comes from, so you can check the claim rather than take it on trust, and carries the standard's suggested fix. The fix button takes you to the field or the cell that needs the answer, not merely to the file.
It also reports things most tools miss, because they are invisible one file at a time: a perfectly named file sitting in a folder that is not a datatype, an entity the standard does not allow for that kind of file, a sidecar left behind next to no data file at all.
Every edit is recorded in the project, and every scan is kept as a version. Undo a decision taken in a session weeks ago, or reopen last month's scan and convert it again against the same answers. Curation is resumable rather than something you redo from the raw files each time.
| Read from | Converted by | |
|---|---|---|
| MRI | DICOM | dcm2niix |
| PET | DICOM | dcm2niix |
| PET | ECAT7, detected by its header rather than a .v name |
nibabel |
| PET blood | PMOD .bld |
pet2bids |
| EEG / iEEG | EDF, BDF, BrainVision, EEGLAB, MEF, NWB | mne-bids |
| EEG / iEEG | Neuroscan, GDF, EGI and other non-BIDS formats | mne-bids, re-encoded to EDF |
| MEG | FIF, CTF .ds, KIT .con/.sqd |
mne-bids |
| Physio | Siemens CMRR log, written beside an MRI series | bidsphysio (vendored) |
Karel López Vilaret and Jochem Rieger, ANCP Lab, Carl von Ossietzky Universität Oldenburg.
MIT.
Physio conversion code under bidsmgr/vendor/bidsphysio/ is derived
from bidsphysio by Pablo
Velasco and Chrysa Papadaniil (NYU Center for Brain Imaging), used
under the MIT License. See bidsmgr/vendor/bidsphysio/LICENSE and
bidsmgr/vendor/README.md for the full attribution and what
changed during vendoring.
López Vilaret, K. M. and Rieger, J.
BIDS Manager (v1.2.6). 2026. https://github.com/ANCPLabOldenburg/BIDS-Manager
Documentation · What changed · Report a bug · Suggest a feature