Results So Far

Introduction to Results

The graphs below show my first results from the current iteration of the software. They reflect the success — or otherwise — of generating Mu and Beta suppression using motor imagery.  After the results graphs, I describe the process of detecting Mu and Beta suppression.  But first the graphs.

As the first chart makes clear, this isn’t a spectacular success. I believe the most immediate issue is the calibration process: getting a stable baseline against which lateralisation can be measured reliably.  We also graph Lateralisation Index (LI) values during each session. These aren’t shown here, but they help with tuning the system and checking whether the balance of rhythms shifts in the expected direction.

I’ll present the overall Success Trend graph first, then discuss how calibration affects the results — and why it’s central to interpreting what the system sees.

How Suppression Is Detected

The goal of this system is to detect Mu and Beta suppression — a drop in specific brain rhythms when someone imagines movement. That’s the core signal we’re looking for.

1. Getting Band Power

We use BrainFlow to measure the power of Mu and Beta rhythms at key electrodes. This gives us raw values for each side of the head — but they’re only meaningful when compared to a baseline.

 

2. Calibration

Before each session, we record a resting baseline. Calibration gives us this baseline of relaxed rhythms. It shows what “normal” looks like when the brain is relaxed.

But calibration is tricky. If the baseline is too low or unstable:

  • Later signals may look artificially strong

  • Suppression may be missed entirely

  • Scores may drop even when the brain is engaged

To fix this, we apply a minimum floor and reject calibration values that are too noisy or implausible. This helps ensure that suppression is measured against something realistic.

 

3. Detecting Suppression

During imagery, we compare live band powers to the calibration baseline.

  • In this system, a drop of around 20% compared to baseline is treated as suppression.

  • This threshold is adjustable and may vary across studies or individuals, but it provides a practical marker for detecting motor imagery.

  • We check the contralateral electrode:

    • Left prompt → check right side (C4)

    • Right prompt → check left side (C3)

    • Centre prompt → both sides should stay close to baseline, with balance near zero

We also run sanity checks to make sure the signal hasn’t collapsed — which would suggest noise, not real suppression.

 

4. Lateralisation: A Supporting Signal

We calculate a Lateralisation Index (LI) to see if the balance shifts in the expected direction:

  • Left prompt → right side should be weaker

  • Right prompt → left side should be weaker

  • Centre prompt → balance should stay near zero

This helps confirm that suppression isn’t happening equally on both sides — which could be misleading.

 

5. Feedback and Scoring

Visual feedback uses percentages to animate balance — expanding or shrinking circles. But scoring uses the hybrid rule:

  • Suppression ratio + Lateralisation direction

  • Only plausible, directional suppression counts as success

Scores are averaged across packets and prompts to reflect consistent patterns — not single spikes.