2D EEG Rhythmicity: Redrawing the Map of Brain Waves

For a century, neuroscience has operated on a deeply entrenched assumption: brain electrical signals fall into five frequency bands — Delta, Theta, Alpha, Beta, Gamma — each with its own role. The model is simple and useful, but it has a blind spot: it cannot explain brief, abrupt, discontinuous anomalous brain signals.

A team from the University of Cambridge and the Hebrew University of Jerusalem recently dropped a bombshell. Using two new algorithms (LAVI and ABBA), they expanded the EEG spectrum from a one-dimensional frequency axis to a two-dimensional frequency–rhythmicity plane, redefining how the brain works.

---

1. The Blind Spot of the Traditional Model

The traditional EEG model does one thing: slice signals by frequency. 8–13 Hz is Alpha, 13–30 Hz is Beta, and so on. Each band maps to a "state" — Alpha for relaxation, Beta for focus, Gamma for higher cognition.

But over the past decade, evidence has pointed to an awkward fact: many brain oscillations are not continuous but burst-like — appearing suddenly, lasting a few hundred milliseconds, then vanishing. These bursts can be as strong as sustained oscillations in the power spectrum, yet they are fundamentally different.

Traditional power spectral density (PSD) analysis cannot tell the difference. It only measures "how strong is this frequency," not "how regular is this frequency."

---

2. LAVI: Phase, Not Power

The LAVI (Lagged-Angle Vector Index), proposed by Golan Karvat's team, rests on a simple idea: rhythmicity is not power, but phase persistence.

Concretely:

• Take the complex signal at a given frequency (via wavelet or Hilbert transform)
• At each time point, compute the phase difference between the current phase and the phase lagged by one cycle
• Sum these phase-difference vectors
• If phases are highly consistent (true rhythm), the vector sum is large; if random (noise or bursts), it is small

LAVI outputs a rhythmicity profile: frequency on the x-axis, rhythmicity (0 to 1) on the y-axis. This profile can be decoupled from the traditional power spectrum — one frequency may have high power but low rhythmicity (a classic burst), another moderate power but extremely high rhythmicity (a classic sustained oscillation).

---

3. ABBA: Letting the Machine Draw the Boundaries

With a rhythmicity profile, the next question is: how to automatically partition bands?

Traditionally, boundaries are hand-picked (e.g., Alpha = 8–13 Hz), but brains differ across individuals and states. Hard-cutting loses information.

ABBA (Automated Band-Border detection Algorithm) works by: 1. Finding significant peaks (high rhythmicity) and troughs (low rhythmicity) in the LAVI profile 2. Using permutation-based statistics to determine which peaks/troughs are real vs. noise 3. Automatically outputting personalized band boundaries for each person, each state

Key point: individual-level automated analysis — no universal template; the data speaks for itself.

---

4. A 2D Architecture: Sustained vs. Episodic

This is the paper's core finding.

The team analyzed 859 participants across 12 independent datasets, spanning:

• Multiple species (human, rat)
• Multiple recording methods (scalp EEG, intracranial electrodes, MEG)
• Ages 18–88
• Multiple regions (visual cortex, hippocampus, prefrontal cortex)
• Health and disease states (including Parkinson's disease)

The result was strikingly consistent:

The brain's spectrum is split by two rhythmicity boundaries into two classes of bands:

| Type | Rhythmicity | Character | Functional role | |------|-------------|-----------|-----------------| | High-rhythmicity bands | High | Sustained, stable phase synchronization | Maintain ongoing activity | | Low-rhythmicity bands | Low | Transient, burst-like | Respond to external input/change |

Specifically:

• Theta, Alpha, Beta1 (low Beta), Gamma1 → high rhythmicity, sustained oscillations
• Beta2 (high Beta), Gamma2 → low rhythmicity, episodic bursts

The traditional model treats Beta as one entity (13–30 Hz). But the 2D model shows low Beta (~13–20 Hz) is sustained while high Beta (~20–30 Hz) is episodic. They are not the same thing.

---

5. Functional Validation: Two Modes, Two Jobs

If high-rhythmicity bands "maintain" and low-rhythmicity bands "respond," the hypothesis should be experimentally testable.

1. Eyes-open vs. eyes-closed

With eyes closed at rest, external input is minimal and the brain mainly maintains internal state. Then:

• Alpha rhythmicity increases significantly (p < 0.01)
• Beta1 (low Beta) rhythmicity also increases
• Beta2 (high Beta) shows no change

As predicted: the maintenance mode strengthens at rest; the response mode is unaffected.

2. Neuronal spiking vs. oscillations

If sustained oscillations "maintain," they should relate to ongoing spiking; if bursts "respond," they should relate to event-driven firing.

The data support this:

• High-rhythmicity bands correlate with baseline firing rate
• Low-rhythmicity bands correlate with stimulus-evoked firing

---

6. Clinical Implications: The Parkinson's Example

Parkinson's patients show abnormally elevated Beta-band power. The traditional model says "too much Beta."

The 2D model asks a more precise question: is it low Beta (sustained) or high Beta (episodic) that goes wrong?

The paper includes Parkinson's data (patients in one dataset); 2D analysis can reveal:

• Elevated sustained Beta → possibly excessive synchronization of motor circuits (classic parkinsonian pathology)
• Elevated episodic Beta → possibly abnormal event responses of the motor system

This distinction has direct implications for precision diagnosis and targeted therapy. For deep brain stimulation (DBS) parameter tuning, knowing which Beta class is affected allows targeted adjustment of stimulation frequency and patterns.

---

7. Why This Matters

The value is not just "two new algorithms" — it changes the basic framework for viewing brain electrical activity:

From "1D" to "2D"

Traditional: frequency → state New framework: frequency + rhythmicity → mode (maintenance vs. response)

From "universal template" to "individualization"

ABBA automatically computes each person's band boundaries. Your Alpha might be 9–12 Hz; mine 8.5–11.5 Hz. One-size-fits-all is replaced by individual precision.

From "power decides everything" to "rhythmicity reveals essence"

Power spectra are only appearance. The same power could be a sustained oscillation (internal maintenance) or a burst (external response). The rhythmicity dimension separates them.

---

8. Limitations and What's Next

The paper is a preprint (bioRxiv 2024.12.05.627113), not yet peer-reviewed.

LAVI and ABBA currently exist only in MATLAB (requiring the Curve Fitting Toolbox); a Python version is not yet released.

The biggest open question: does this 2D architecture hold for all brain regions, all cognitive states, all diseases? 859 subjects and 12 datasets is large, but sample diversity in neuroscience is never exhausted.

---

One-Sentence Summary

The traditional EEG model is a map with only a horizontal axis, labeling "Alpha here, Beta there." LAVI and ABBA add a vertical axis — rhythmicity — revealing two continents: a land of sustained oscillations (maintenance) and a sea of episodic bursts (response). Brain electrical activity now has two-dimensional coordinates.

Preprint: https://doi.org/10.1101/2024.12.05.627113 Toolbox: https://github.com/laaanchic/LAVI