Gradient Biotech

Pipelines

Every analysis step you run in the app maps to one of these pipelines. This reference lists what each one takes as input and produces as output.

Connectivity and graph theory

PipelineInputOutputs
graph_metricsConnectivity matrix (CSV, NPZ, MAT)Global BCT metrics, per-node metrics, module assignments
graph_sensitivityConnectivity matrixDensity-sweep robustness report: BCT metrics across thresholds + per-metric coefficient of variation / stability flag

Key parameters: threshold (edge weight cutoff), n_rand_smallworld (null model count).

EEG

PipelineInputOutputs
eeg_preprocessEDF, BrainVision, EEGLabFiltered recording, bad channels, optional ICA summary
eeg_featuresEEG preprocess runBand power, spectral entropy, PLV/coherence/wPLI connectivity
eeg_artifactsRaw EEG (EDF, BrainVision, EEGLab, FIF)Per-channel QC metrics, ICA component maps/timecourses/spectra, auto EOG/muscle flags, raw preview
eeg_connectivityEEG preprocess run or raw EEGMulti-estimator connectivity for a band (coherence, imaginary coherence, PLV, PLI, wPLI, debiased wPLI, PPC, AEC) + estimator-agreement matrix
ieeg_analysisIntracranial EEG/ECoG (EDF, FIF, BrainVision)Shaft-aware bipolar referencing, band power incl. high gamma, peri-event high-gamma response
eeg_apply_cleaningRaw EEG + reviewed ICACleaned recording with confirmed bad-channel interpolation and selected ICA components removed; before/after variance
microstatesEEG preprocess runState maps, coverage, duration, transitions, GEV
time_frequencyEEG preprocess runSpectrogram stats, band-power dynamics
source_localizeEEG preprocess runSimplified source-space summary

fMRI

PipelineInputOutputs
bids_validateBIDS ZIP archiveValidation report, missing-field warnings
fmri_qc4D fMRI NIfTI (+ optional confounds/motion file)Framewise displacement, DVARS, outlier scrubbing, tSNR, subject-exclusion flag
task_glm4D fMRI NIfTI + BIDS events (onset, duration, trial_type)First-level ROI GLM: HRF design matrix, per-condition beta/z, contrast effect/t/z, event-timing QC
functional_connectivityPreprocessed NIfTIROI timeseries, Pearson correlation matrix, Fisher z
dynamic_fcFC matrix or NIfTISliding-window states, state timecourses

Key FC parameters: atlas_name (aal, schaefer_100/200/400), confound_strategy, bandpass cutoffs.

Cohort

PipelineInputOutputs
graph_metrics_batchMultiple matrix datasets with group labelsPer-subject metrics table
cohort_nbsGroup-labeled connectivity matricesNBS components, edge statistics, heatmap data
cohort_metricsBatch graph metrics resultsGroup comparison table with effect sizes
cohort_predictGroup-labeled or target-valued connectivity matricesCross-validated classification (ROC AUC, balanced accuracy, sensitivity/specificity) or regression (R², MAE, Pearson r), permutation p-value, edge-stability map, subject-level predictions
cluster_statsGroup-labeled EEG features runsSensor × band cluster permutation test (cluster mass, max-statistic FWE p-values), t-map, significant clusters
multiplex≥ 2 same-node connectivity matrices (layers)Multiplex metrics: per-layer density/strength, overlapping strength, participation coefficient, layer-similarity matrix

Interpretation

PipelineInputOutputs
interpretCompleted metrics / NBS / prediction / microstate runNarrative JSON with metric citations

Workflow templates

Reusable protocol presets chain the pipelines above so you can run a multi-step protocol from one panel instead of launching each pipeline individually. The Run protocol panel executes the chain one step at a time, carrying each step's output into the next.

TemplateModalitySteps
Resting EEG → microstatesEEGeeg_preprocesseeg_featuresmicrostates
EEG connectivityEEGeeg_preprocesseeg_features
Resting-state fMRI → dynamic FCfMRIfunctional_connectivitydynamic_fc

Job lifecycle

Every run moves through the same states, shown as a status badge on the dataset and cohort pages:

queued → running → complete | failed

Versioning

Every run records a pipeline version. Include run IDs and versions in methods sections for reproducibility.

Implementation sources

Graph metrics are ported from Brain Connectivity Toolbox (BCT) reference algorithms. EEG preprocessing uses MNE-Python. Functional connectivity uses nilearn atlas parcellation.