Preprocess Signal

Prepare derived signal state without rewriting raw EDF.

Preview bad-channel, drift, re-reference, ASR, and ICA workflows locally before using processed sources in analysis.

Preprocessing in EEG TraceLab is designed as a local research preparation workflow. Raw EDF samples remain immutable; processed runs are derived local signal state with warnings, parameters, and provenance.

Raw source preserved

Preprocess Signal

Bad channelsDriftASRICA
Raw stays rawImported EDF is not overwritten
ASR / ICALocal preview workflows
ProvenanceParameters and warnings exported

Prepare workflow

Manual review first, signal-quality suggestions second.

Users can mark bad channels manually with notes and reasons, then review suggested bad channels as signal-quality cues before applying decisions.

Manual bad channel marking, unmarking, clear actions, notes, and reasons
Suggested bad channels remain separate until user confirmation
Flatline, extreme variance, line noise, clipping, invalid/dropout samples, neighbor correlation, kurtosis, and amplitude range cues
10-20 / 10-10 scalp label recognition with unknown and ambiguous label warnings

Preview effects

See what preprocessing choices do before saving a run.

Preprocess Signal exposes before/after context for drift, detrending, high-pass preview, re-reference choices, ASR, and ICA.

Drift and filtering

Drift scoring, linear detrend, polynomial detrend, moving-median detrend, optional high-pass preview, before/after trace, PSD comparison, and parameter recording.

Re-reference

Preview keep current/no change, common average, selected reference channels, and linked/custom reference pairs when labels exist.

ASR and ICA

Advanced local previews with warnings attached.

ASR and ICA are presented as cautious local research workflows, not automatic cleaning or interpretation.

ASR preview

Clean calibration segment suggestion, manual calibration selection, calibration covariance, sliding-window covariance/PCA high-variance subspace identification, affected-window table, reconstructed component counts, and rank warnings.

ICA preview

Deterministic FastICA-style decomposition, component time series, spectra, variance contribution, component topography when positions exist, manual component exclusion, convergence diagnostics, and preview reconstruction.

Raw vs Processed

Analysis can use immutable raw samples or saved processed runs.

Changing the data source marks existing results dirty and asks for Compute / Recompute before long processed-source work starts. Cancel keeps previous results visible and does not create a processed run or mutate raw EDF data.

Preprocessing Runs and Preprocessing Presets in Workspace
Built-in Scalp EEG review, SEEG review, Infraslow-safe, Fast preview, ASR conservative review, and ICA manual component review presets
CSV / JSON / PNG / bundle preprocessing report export where eligible
Infraslow is valid only with true DC-coupled recordings.
No preprocessing result should be presented as clinical cleaning, automatic artifact diagnosis, or medical interpretation.