Outputs

Outputs#

MEGFlow writes Nextflow execution files under params.megflow.output_dir and processing derivatives under the dataset preproc_dir. By default:

preproc_dir = <dataset_output_dir>/preprocessed

Top-Level Output Layout#

Path	Description
`<output_dir>/work/`	Nextflow work directory.
`<output_dir>/static_html_report/nextflow/report.html`	Nextflow execution report for a single-dataset run.
`<output_dir>/static_html_report/nextflow/timeline.html`	Nextflow timeline for a single-dataset run.
`<output_dir>/static_html_report/nextflow/trace.txt`	Nextflow process trace. The static report uses it to populate per-subject `Task Details` when available.
`<output_dir>/static_html_report/nextflow/nextflow.log`	Nextflow driver log when the launcher supplies the documented `-log` path.
`<output_dir>/nextflow.config`	Runtime config copied by the distributed Docker entrypoint. Source launches do not create this file automatically.
`<output_dir>/static_html_report/`	Portable MEGFlow QC report. Depending on `static_task_log_mode`, this may include packaged Nextflow `.command*` log excerpts.
`<output_dir>/preprocessed/`	MEGFlow processing derivatives.

Corpus Output Layout#

With Docker --corpus or a source config that sets corpus_root, every dataset receives an isolated output tree. Dataset names are sanitized for use as directory names, while the corpus report retains the configured display name.

<output_dir>/
|-- datasets/
|   |-- <dataset_a>/
|   |   |-- preprocessed/
|   |   `-- static_html_report/       # MEG/report stages
|   `-- <dataset_b>/
|       |-- preprocessed/
|       `-- static_html_report/       # MEG/report stages
|-- corpus_static_html_report/
|   |-- index.html
|   |-- assets/
|   |-- data/
|   |   |-- corpus_summary.json
|   |   `-- datasets.csv
|   |-- datasets/                     # bundled portable dataset reports
|   `-- nextflow/
|       |-- nextflow.log
|       |-- report.html
|       |-- timeline.html
|       `-- trace.txt
`-- nextflow.config

corpus_static_html_report/index.html is the cross-dataset entry point. It links the dataset reports and provides sortable recording and dataset summaries for comparison, outlier discovery, and prioritizing manual review. The files under corpus_static_html_report/nextflow/ describe the complete Nextflow invocation rather than one dataset. They are stored once and linked from the corpus report instead of being copied into every dataset report. An anatomy-only dataset has derivatives and reconstructed anatomy but no dataset-level static MEG report unless a later MEG or report stage creates one.

Preprocessed Directory#

Path	Contents
`preprocessed/<recording>/`	Continuous preprocessed raw files, ICA-clean raw files, and selected QA plots.
`preprocessed/quality_control/<recording>/`	NormMEG-QC summary JSON, component-score CSV, and NMDQ score figure when `megqc.enabled` is true.
`preprocessed/artifact_report/<recording>/`	Bad-channel files, detector provenance, bad-segment annotation files, DeepReject provenance when enabled, a recording-wide mask heatmap, and optional detailed artifact review images.
`preprocessed/ica_report/<recording>/`	ICA model, component labels, ECG/EOG score files, component figures, and overlay/PSD plots.
`preprocessed/epochs/<recording>/`	Epoch FIF files, rejection logs, and epoch-level figures.
`preprocessed/covariance/<recording>/`	`bl-cov.fif` and diagnostics; conditional `lcmv-data-cov.fif` and diagnostics when LCMV is selected; and the always-present `resolved-rank.json` source contract.
`preprocessed/trans/<recording>/`	MEG-MRI transform, coregistration distance CSV, and staged coregistration figures.
`preprocessed/forward_solution/<recording>/`	Forward model outputs and head model figures.
`preprocessed/source_recon/<recording>/`	Source reconstruction outputs and visualization figures.
`preprocessed/logs/`	`megflow_run_manifest.json`. Nextflow logs and execution reports live under the report package’s `nextflow/` directory; the Docker runtime config is copied to `<output_dir>/nextflow.config`.
`preprocessed/deepprep/`	DeepPrep outputs when `anatomy.method = "deepprep"` and anatomy processing is enabled.
`preprocessed/pseudomri/<subject>/`	Generated pseudo T1 input when `anatomy.method = "pseudomri"`.

FreeSurfer-format anatomy is written to the configured fs_subjects_dir. In a corpus run without an explicit override this is normally under <output_dir>/smri/<dataset_name>/; it is not a recording derivative.

Important Sidecar Files#

File pattern	Meaning
`*_preproc-raw.fif`	Continuous output from OSL preprocessing.
`*.summary.json`	NormMEG-QC NMDQ score, score metadata, family scores, and processing-gate status.
`*.component_scores.csv`	NormMEG-QC component values and reference-calibrated subscores.
`*.normative_quality_score.png`	NMDQ score and metric-family score figure.
`*_bad_channels.txt`	One bad channel name per line.
`*_bad_channels_description.json`	Per-channel detector provenance showing which enabled method marked each final bad channel.
`*_bad_segments.txt`	MNE annotation file containing bad time spans.
`deepreject_summary.json`	DeepReject input preprocessing, folds, thresholds, channel probabilities, bad-channel decisions, bad intervals, and runtime settings. It is written only when the DeepReject branch runs successfully.
`check_imgs/artifact_mask_heatmap.jpg`	Whole-recording channel-by-time mask overview. This compact image is generated even when `artifact_images_enabled` is false.
`*_ica.fif`	Fitted ICA model.
`marked_components.txt`	Final ICA component indices selected for exclusion. In automatic mode this is the union of enabled ECG, EOG, and outlier categories; a preserved manual review may replace that automatic union.
`ecg_eog_scores.json`	Enabled-category `ecg_indices`, `eog_indices`, and `outlier_indices`, the resolved `category_switches`, and method provenance. Its `marked_components.auto_indices` records the automatic category union, while `written_indices` exactly matches `marked_components.txt`.
`*_clean_raw.fif`	Continuous raw file after ICA application.
`*-epo.fif`	Epoch output.
`*_reject_epoch_log.txt`	Rejected epoch indices and estimated remaining epoch count.
`bl-cov.fif`	Noise covariance estimate.
`lcmv-data-cov.fif`	LCMV data covariance from the exact source Raw/Epochs. This file is not generated for minimum-norm-only runs.
`resolved-rank.json`	Resolved target-rank dictionary, ordered common-channel list, and source input mode shared by covariance and source reconstruction.
`dists.csv`	Coregistration distance summary in mm.
`coreg-trans.fif`	MEG-MRI transform used by forward modeling.
`megflow_run_manifest.json`	Pipeline mode, selected stages, path snapshot, and runtime metadata.

The exact set of outputs depends on the effective steps value in params.megflow. For example, meg_artifacts does not produce ICA, epoch, covariance, forward, or source outputs, while report only rebuilds the static report from existing files.

Artifact Image Policy#

artifacts.artifact_images_enabled controls detailed waveform and detector review figures; it does not disable the compact artifact-mask heatmap. The heatmap gives reports a consistent recording-wide summary without incurring the cost of all detailed figures. artifact_image_n_jobs controls parallel image generation only and does not change artifact decisions. Its automatic value, and any explicit upper bound, cannot exceed the detect_artifacts CPU allocation.

Bad-segment sidecars contain annotations, not shortened data. Samples remain in the continuous FIF. The normal ICA path carries them into the cleaned raw; the meg_epochs,skip_ica path loads the bad-channel and bad-segment sidecars into the preprocessed raw immediately before epoching. ICA fitting and epoch construction exclude annotated spans according to their own reject_by_annotation behavior. See Bad Segments: Marking vs Exclusion for the exact distinction between marking and later exclusion.