DEBUG LOG CHAMPAGNE eLORETA
EDF · CNT · CBT · CSV · AVG · FDT · BIN · VHDR
← Back
No file loaded
MOVIE PLAYBACK
SUBJECT VIDEO
Fit Diagnostics
MNR ANALYSIS INTERVAL
Type the interval in milliseconds from the start of the loaded data. For averaged ERP files this is relative to the stim onset baked into the file.
Start (ms)
Stop (ms)
 
Acquire
Scroll View
Epoch View
Epoch Browser
Topo Map
Spectral
Champagne
Multi-Dipole Fit
MNR
Simulate
MRI
Grand Avg
Visibility
SCALE
No EEG Data Loaded
Drop a CNT / CBT / CSV / AVG file or load a demo
Loading EEG Data
Reading file…
POS
0.00s / 0.00s
ERP VIEWER
Amplitude Scale
5µV
Analysis
Source localization → use the MNR / Multi-Dipole Fit / Champagne tabs (press ← Back first to bring this average into the workspace).
Shift+drag to select
a time window first
Actions
Channels
Labels
Hover — cursor readout
Click — mark peak
Shift+drag — select window
Dbl-click — clear all
AVG / ERP
EPOCH SETTINGS
Pre-stimulus ms
Post-stimulus ms
Baseline ms pre
Set intervals and click Extract to epoch around markers.
+ Add marker here
+ Add marker at cursor
✕ Reject this epoch
Clear all markers
ADD MARKER
Color:
Source Localization Validation
Truth model (simulates)
Test models (invert)
Montage
Noise (SNR dB) — Inf
Repeats / source (noise avg)
Known dipoles are placed across a sweep of regions (occipital, temporal, central, frontal) and eccentricities (0.3 / 0.5 / 0.7 of head radius), each with an oblique moment. The truth model generates scalp potentials; each test model inverts them via grid-scan with an analytic least-squares moment solve. Error is the distance between the true and recovered dipole (head radius assumed 85 mm). When truth = test, this is the optimizer/grid floor (“inverse crime”); when they differ, it measures model-mismatch error.
Compare Source Methods
Every inverse method here runs on the same averaged ERP over the same interval (taken from the MNR Start/Stop ms, mirrored into the other engines for this run; each engine keeps its own forward model): MNR, Champagne, eLORETA, SSLOFO, CLARA and the Multi-Dipole Fit. LORETA has no compute entry point of its own on the source panel and sLORETA runs from its own tab, so neither is included here rather than being listed and quietly stood in for. All but one are distributed — their estimate is the peak voxel; Multi-Dipole Fit returns discrete dipoles (strongest shown). On real data there is no ground truth, so the comparison reports how much the methods agree (pairwise distance between peaks, consensus location, anatomical concordance) and each method's own fit-quality diagnostics (Multi-Dipole residual variance; iteration count, convergence and focality for each distributed method; MNR focality and spread). If a simulated source is loaded, localization error vs ground truth is added. Use “Save report” to export these statistics. Each engine's full map stays on its own tab.

Σ Grand Average

no files selected
Missing channels
Time-base
|amp| p2p
Baseline
Weighting
Combine
▸ Advanced filters & output

σ² TANOVA — Topographic ANOVA

Assign each file to condition A or B (or skip). Maps are average-referenced and normalized to GFP = 1, then averaged over the window below; significance of the topographic difference (GFP of the A−B difference map) is obtained by label randomization.
Simulate test set subj/cond
Test
Design
Randomizations
Window start ms
Window end ms
p-threshold
Normalize
▸ PROCESSING STEPS