Cloud
PiEEG Studio (MNE)

PiEEG Studio (Browser & Cloud Preview)

Browser-based MNE-Python EEG analysis and MOABB BCI benchmarking — zero install required

PiEEG Studio (opens in a new tab) is now available directly in the browser on PiEEG Cloud (cloud.pieeg.com). Launch it from the Session Lobby header, click Studio ↗ on any saved session in your Saved Sessions list, or send a freshly captured in-memory browser recording straight from the playback toolbar — without installing Python or MNE locally.

If you also run the Community Edition locally (http://127.0.0.1:1618) on your machine, cloud.pieeg.com automatically detects it (GET http://127.0.0.1:1618/api/health) and routes Studio actions to your local instance for unlimited file sizes, 150+ online MOABB datasets, and native Qt desktop MNE browsers.

PiEEG Studio Overview

Open PiEEG Studio → pieeg-studio.fly.devOpen PiEEG Cloud → cloud.pieeg.comSelf-Hosted / Local CLI Guide

Cloud Preview vs. Local Community Edition

CapabilityBrowser Cloud Preview (pieeg-studio.fly.dev)Local Community Edition (127.0.0.1:1618)
InstallationNone — runs in any browserpip install "pieeg-server[studio,studio-moabb]"
Auto-selection from cloud.pieeg.comDefault fallback when local port 1618 is not runningAutomatically preferred when reachable on 127.0.0.1:1618
Session & data isolationStrict per-browser X-Studio-Session (ps-sid) + 15-min idle purge100% local disk & memory
Direct Cloud session loaderYes (?cloudSession=<id> + event markers)Yes
In-memory recording handoffYes (POST /api/datasets CSV blob + annotations)Yes
Max upload / dataset quota25 MB per file · up to 4 active datasets per sessionUnlimited (RAM-bound)
MOABB BCI Benchmark LabBuilt-in offline synthetic MI / P300 / SSVEP + open session tabs+ 150+ online public NeuroTechX MOABB datasets & custom disk cache
Native desktop MNE windowsBrowser MNE Visual Lab (PNG) + .ipynb export for Colab/KaggleInteractive Qt (raw.plot()) & matplotlib desktop windows

Launching Studio from PiEEG Cloud

From the Session Lobby (cloud.pieeg.com/lobby)

Click Studio in the top navigation bar (or More → PiEEG Studio) to open your isolated Studio workspace in a new tab. You can drag-and-drop any local .csv, .xlsx, .edf, .bdf, .gdf, .fif, .vhdr, or .set recording, or click Try a demo recording to generate a 16-channel synthetic session with eyes-open/closed blocks, blinks, and 50 Hz mains noise.

From Saved Sessions

Open Saved Sessions in the dashboard and click Studio ↗ next to any cloud recording. Studio downloads the session CSV (/v1/sessions/{id}/download) and its cloud event markers (/v1/sessions/{id}/events) directly into MNE-Python.

From an Unsaved In-Memory Recording (SessionViewer)

Right after capturing a recording in your browser — even before publishing or signing in — click Studio ↗ in the playback toolbar. The browser uploads the in-memory CSV Blob and your frame annotations directly to your isolated Studio session (POST /api/datasets) with 10–20 montage mapping enabled.

Automatic local handoff: Whenever you click Studio or Studio ↗ in cloud.pieeg.com, the client probes http://127.0.0.1:1618/api/health (350 ms timeout). If you have pieeg-studio running locally, it opens your local instance; otherwise it launches the hosted Cloud Preview at https://pieeg-studio.fly.dev.

Privacy & Per-User Isolation

Because the Cloud Preview runs on shared infrastructure, every browser session is strictly sandboxed:

  1. Per-browser Session ID (ps-sid): Your browser generates a persistent UUID (localStorage["ps-sid"]) sent via the X-Studio-Session header. All in-memory MNE Dataset objects, cached figures, and CSP patterns are keyed to your sid — other users can never list or access your recordings.
  2. Zero-leak token transport: When opening a private saved session from cloud.pieeg.com, your Cloud access token is passed in the URL hash (#sid=...&token=...) so it is never sent in HTTP Referer headers or server access logs, and Studio strips the hash from the address bar immediately after loading.
  3. 15-minute inactivity reaper & scale-to-zero: A background reaper purges any session's datasets, temporary files, and cached plots after 15 minutes of inactivity (PIEEG_STUDIO_IDLE_TTL_S=900), and the Fly.io machine automatically stops when idle.

Analysis & Preprocessing Views

PiEEG Studio in the browser provides the full 11-view MNE-Python workflow:

ViewHighlights
OverviewRecording summary, sampling rate & 10–20 montage check, timing-gap diagnostics, quick PSD & channel preview
Signal browserMulti-channel WebGL viewer (scattergl), min/max overview minimap, raw vs. processed overlay, DC removal toggle, and click-to-annotate markers
Channels & qualityPer-channel std, peak-to-peak, RMS, 50/60 Hz mains noise %, kurtosis, Hjorth parameters (activity, mobility, complexity), and bad-channel flagging
PreprocessingReorderable live pipeline: DC offset / linear detrend, centered Hampel spike filter, FIR/IIR notch (+ harmonics), FIR/IIR band-pass, bad-channel spherical-spline interpolation, CAR / custom re-reference, resampling, cropping, and ICA — with live before/after PSD comparison and mne.viz.plot_filter response
Artefacts & ICAAutomatic blink (frontal EOG proxy), amplitude (BAD_peak), and muscle (BAD_muscle) detectors; FastICA / Extended Infomax / Picard decomposition with automatic ocular component suggestions, component topomaps, properties, and source browser
Spectral & bandsWelch / multitaper PSD, $\delta / \theta / \alpha / \beta / \gamma$ absolute & relative band powers, band-power ratios ($\theta/\beta$, $\alpha/\theta$, $\beta/(\alpha+\theta)$), IAF, SEF95, spectral entropy, $1/f$ slope, frontal alpha asymmetry ($\ln R - \ln L$), and mne.viz.plot_topomap band maps
Time–frequencyPer-channel short-time spectrogram and whole-session stacked band-power timeline
Events & ERPEpoching around annotations or fixed windows, baseline correction, peak-to-peak rejection, condition averages $\pm$ SEM, GFP, evoked.plot_joint, and epochs.plot_image
ConnectivityCoherence, PLV, wPLI, and band-limited correlation rendered as an interactive adjacency matrix and MNE circular connectivity diagram
MNE visual labPublication-ready MNE figures (raw.plot, compute_psd().plot, plot_sensors, plot_topomap, plot_filter, ica.plot_components) + live syntax-highlighted MNE Python code
BCI & MOABBStandardized BCI benchmark lab comparing 7 spatial/Riemannian decoding pipelines with Stratified K-Fold CV, ROC-AUC, confusion matrices, and CSP pattern topomaps
Export & code.xlsx (6-sheet workbook), .csv, .fif, .edf, .html (mne.Report), .py, .ipynb (Colab/Kaggle-ready Jupyter notebook), .json features, and reusable pieeg_pipeline.json presets

BCI & MOABB Benchmark Lab

The BCI & MOABB tab brings MOABB (Mother of All BCI Benchmarks) (opens in a new tab), MNE epoching, and Riemannian geometry decoding directly into PiEEG Studio:

  • Synthetic & Recorded Paradigms: Generate instant offline synthetic datasets (Motor Imagery, P300 / ERP, SSVEP) subsetted to PiEEG channel layouts (PiEEG-8, PiEEG-16, IronBCI-32, or all channels), or benchmark your own annotated recordings open in Studio tabs (Within-Session CV or Cross-Session across open tabs).
  • 7 Built-in Decoding Pipelines:
    1. CSP + LDA — Common Spatial Patterns + Shrinkage LDA
    2. CSP + SVM — Common Spatial Patterns + RBF SVM
    3. Riemannian TS + LR — SPD Covariance + Riemannian Tangent Space + Logistic Regression
    4. Riemannian MDM — Minimum Distance to Riemannian Mean
    5. LogVar + LDA — Channel Log-Variance + Shrinkage LDA
    6. BandPower + RF — Welch Band Powers + Random Forest
    7. ERP-Xdawn + LDA — ERP Spatial-Temporal Covariance + LDA
  • Visual Diagnostics & Reproducible Script: Compare pipeline Accuracy, Balanced Accuracy, F1, and ROC-AUC against chance level, inspect per-pipeline confusion matrices and MNE CSP spatial pattern topomaps (csp_patterns.png), and download the standalone pieeg_moabb_benchmark.py script.

| Environment Variable | Default | Description |
|---|---|---|
| `PIEEG_CLOUD_URL` | `https://pieeg-cloud.fly.dev` | PiEEG Cloud API base URL for loading saved sessions and event markers |
| `PIEEG_STUDIO_IDLE_TTL_S` | `900` | Seconds of inactivity (15 min) before a browser session's datasets are purged |
| `PIEEG_STUDIO_MAX_DATASETS` | `4` | Maximum concurrent open datasets per browser session (`X-Studio-Session`) |
| `PIEEG_STUDIO_MAX_UPLOAD_MB` | `25` | Maximum file upload / cloud session download size in MB |
| `VITE_STUDIO_URL` *(frontend)* | `https://pieeg-studio.fly.dev` | Custom Studio URL used by `cloud.pieeg.com` |
Live Lab (Bio-Pen)Buddy AI