WaveFM · EEG foundation model

Clean EEG for every analysis.

WaveFM, our EEG foundation model, finds and labels artifacts in every channel and time window. Your analysis starts from clean data, and we can run the qEEG analysis too.

See how it works → Contact us
example_recording.edf Artifacts labeled
Fig. 1 Artifacts labeled by type, per channel and time window. Illustration, not patient data or model output.
  • Eye
  • Muscle
  • Chewing
  • Shiver
  • Electrode

Partners & programs

NVIDIA Inception Program Google for Startups AWS Startups Seoul National University Bundang Hospital TrustConnector

01 The problem

Every EEG has artifacts. Finding them is slow.

Eye movement, muscle, chewing, shiver and electrode artifacts appear in every recording. Today they are usually marked by hand, one recording at a time, before any analysis can start.

49%

of resting EEG was artifact-free after standard preprocessing, across 14 study sites and 305 people. Poor quality also raised spectral power: it changes the result.

0.54

agreement between two experts marking artifacts in the same recordings (Cohen's kappa, where 1.0 is full agreement).

Sources: Kaiser A et al., EEG Data Quality: Determinants and Impact in a Multicenter Study, Brain Sciences 2021 (48.9% eyes open, 49.0% eyes closed). Diachenko M et al., eNeuro 2022.

02 How it works

Manual review becomes one automatic, consistent step.

Step 1 Upload A standard EEG file, to our cloud Step 2 Label WaveFM marks every channel and window Step 3 Exclude Or repair, if your study needs it Step 4 Export Analysis-ready EEG and a map of every change Step 1 Upload Step 2 Label with WaveFM Step 3 Exclude (or repair) Step 4 Export
Fig. 2 From raw recording to analysis-ready EEG
Artifact detection & qEEG →

Also from Balkeum Labs

Federated learning, analytics and secure computation, for data that can't move. Learn more →

Let's talk about your EEG data.

Tell us about your recordings and your study. We'll tell you how WaveFM can help.

Contact us →