EEG - fNIRS Hybrid Imaging: The Perfect Balance of Temporal and Spatial Precision

EEG Data Loading and Basic Filtering (Updated)

This notebook will help you complete Assignment 1 step by step.
Includes ! xed Step 4 (Frequency Analysis) with both individual electrode
spectra and average spectrum .

PSYCH 403A1 - Assignment 1  Setting up your brain data analysis toolkit... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 7.4/7.4 MB 32.7 MB/s eta 0:00: SUCCESS! All tools loaded and ready to go! MNE version: 1.10.1 NumPy version: 2.0.2 # STEP 1: Install & Import Required Packages print ( " Setting up your brain data analysis toolkit..." ) ! pip install -q mne matplotlib numpy scipy requests import mne import numpy as np import matplotlib.pyplot as plt from scipy import signal import requests import os from pathlib import Path plt.rcParams[ 'figure.figsize' ] = ( 10 , 6 ) plt.rcParams[ 'font.size' ] = 11 plt.rcParams[ 'lines.linewidth' ] = 1.5 mne.set_log_level( 'WARNING' ) print ( " SUCCESS! All tools loaded and ready to go!" ) print ( f " MNE version: {mne. __version__ } " ) print ( f " NumPy version: {np. __version__ } " ) # STEP 2: Load Brain Wave Data print ( " Loading brain wave data (this might take a moment)..." ) 

Loading brain wave data (this might take a moment)...
SUCCESS! Brain data loaded!
Data info:
Length: 277.7 seconds
Channels: 59 EEG electrodes
Sampling rate: 601 Hz
sample_data_folder = mne.datasets.sample.data_path ()
raw_file = Path(sample_data_folder) / 'MEG' / 'sample' / 'sample_audvis_ra
raw = mne.io.read_raw_fif(raw_file, preload= True )
raw.pick_types(meg= False , eeg= True , stim= False , exclude= 'bads' )
print ( " SUCCESS! Brain data loaded!" )
print ( f " Data info:" )
print ( f " Length: {raw.times[ -1 ] :.1f } seconds" )
print ( f " Channels: { len (raw.ch_names)} EEG electrodes" )
print ( f " ⚡ Sampling rate: {raw.info[ 'sfreq' ] :.0f } Hz" )
raw.plot(n_channels= 10 , duration= 5 , scalings= 'auto' )

EEG 001
38.67 V
EEG 002
EEG 003
EEG 004
EEG 005
EEG 006
EEG 007
EEG 008
EEG 009
EEG 010
2

Help

50 100 150 200

50 100 150

Time (s) 200 Applying band-pass filter (1-40 Hz)... Filter applied! # STEP 3: Basic Filtering print ( " Applying band-pass filter (1-40 Hz)..." ) raw_filtered = raw.copy(). filter (l_freq= 1 ., h_freq= 40 .) print ( " Filter applied!" ) raw_filtered.plot(n_channels= 10 , duration= 5 , scalings= 'auto' )

Calculating frequency spectrum...
# STEP 4: Frequency Analysis (Power Spectral Density)
print ( " Calculating frequency spectrum..." )
# Use modern MNE function
psd = raw_filtered.compute_psd(fmin= 1 , fmax= 40 , n_fft= 1024 )
psds, freqs = psd.get_data(return_freqs= True )
# Plot all electrode spectra
plt.figure(figsize=( 10 , 5 ))
for psd_line in psds:
plt.plot(freqs, psd_line, color= 'lightgray' , alpha= 0.5 )
# Add average spectrum (bold)
mean_psd = psds.mean(axis= 0 )
plt.plot(freqs, mean_psd, color= 'red' , linewidth= 2 , label= 'Average Spectru
plt.xlabel( "Frequency (Hz)" )
plt.ylabel( "Power (dB)" )
plt.title( "EEG Frequency Spectrum (1–40 Hz)" )
plt.legend()
plt.show()

Start coding or generate with AI.

Code: We had troubles saving the code so it is pasted below:
// Brain particle system
let particles = [];
function setup() {
// HSB in 0..1 (you already had this)
colorMode( HSB , 1 , 1 , 1 );
for ( let i = 0 ; i < 100 ; i++) {
particles.push({
x: random(width),
y: random(height),
// base fast velocities (your choice kept)
vx: random( 0 , 50 ),
vy: random(- 2 , 50 ),
life: 1.0 ,
seed: random( 1 ) // NEW: per-particle hue for rainbow
});
}
}
function draw() {
// Use BrainImation sliders: Attention≈beta, Meditation≈alpha
const alpha = (eegData.meditation ?? eegData.alpha ?? 0 );
const beta = (eegData.attention ?? eegData.beta ?? 0 );
// stress logic: true if attention > meditation
const stressed = beta > alpha + 0.02 ; // tiny margin avoids flicker
// BG: stressed = white, calm = dark
if (stressed) background( 0 , 0 , 1 , 0.15 );
else background( 0 , 0 , 0.06 , 0.10 );
// speed scale: calm slow, stressed fast
const speed = stressed ? 1.0 : 0.18 ; // <<< slow when calm
for ( let p of particles) {
// Movement influenced by brain waves (your logic kept)
p.vx += (random(- 1 , 1 ) * (eegData.alpha ?? alpha) * 0.1 );
p.vy += (random(- 1 , 1 ) * (eegData.beta ?? beta) * 0.1 );
// Apply speed scaling (NEW)
p.x += p.vx * speed;
p.y += p.vy * speed;
// Wrap around
if (p.x < 0 ) p.x = width;
if (p.x > width) p.x = 0 ;
if (p.y < 0 ) p.y = height;
if (p.y > height) p.y = 0 ;
// Color: stressed = red, calm = rainbow (NEW)
let hue = stressed ? 0.0 : (p.seed + frameCount * 0.002 ) % 1.0 ;
fill(hue, 0.9 , 0.95 , p.life);
noStroke();
ellipse(p.x, p.y, 5 + (eegData.gamma ?? 0 ) * 10 );
// life (your fade kept)
p.life *= 0.995 ;
if (p.life < 0.1 ) p.life = 1.0 ;
}
}

Demo
Calm Mode:
Meditation ≈ 0.8 ↑ Attention ≈ 0.2 ↓
Dark background, slow/rainbow drift
Stressed mode:
Attention ≈ 0.9 ↑ Meditation ≈ 0.1 ↓
White background, red fast streaks

Zen Blood Particles: A Brain-Controlled Visualization of Stress and Calm
Contributors:
Ayaan Jimale, Sarah Badran, Marinaya Saigh, Hamza Razzo
Concept
Our project began with the Brain Particles example from the BrainImation platform.
We wanted to transform a simple particle system into a bio-responsive visualization of stress and calm
— a sort of “digital mood” that reflects brain activity. When a user becomes stressed (higher beta /
attention activity), the screen erupts into fast-moving red particles against a white background,
resembling blood splatter or heightened physiological arousal. When the user relaxes (higher alpha /
meditation activity), the environment shifts to a dark background with slow, softly colored rainbow
particles drifting gently. The contrasting aesthetics represent two opposing brain states: tension versus
calm.
Technical Implementation
We used the built-in BrainImation variables eegData.attention (as a proxy for beta ) and
eegData.meditation (as a proxy for alpha ).
The program continuously compares these two values to determine the current state:
if attention > meditation → stressed mode
else → calm mode
Stressed Mode
● White background
● Red particles
● High velocity (up to 50 px per frame)
● Slight downward/right bias to suggest urgency and gravity
Calm Mode
● Dark background
● Rainbow-cycling particle hues (based on each particle’s seed and frame count)
● Reduced velocity for smoother, slower drift
To avoid flicker during rapid EEG fluctuations, the code includes a small threshold (attention >
meditation + 0.02) before switching states.
This keeps transitions visually stable while remaining responsive in real time.
The system also uses the “Simulate Data” mode for testing, with the Attention and Meditation
sliders controlling stress and calm levels manually.
When connected to a Muse headset, these values update automatically from the user’s EEG signals.
Challenges
Early versions of the code produced constant flickering because simulated EEG data changed
frame-by-frame. We fixed this by slightly smoothing the inputs and introducing a small difference

threshold before toggling modes. Another challenge was balancing aesthetics and performance: large
particle velocities looked dramatic but caused visual clutter, so we tuned the speed scale for a more
natural flow. We also adjusted HSB color ranges so that red remained vivid on a white background
while rainbow hues remained visible on a dark one.
What We Learned
Working with BrainImation showed how small numerical differences in brain-wave bands can create
striking visual differences. We learned how to translate abstract EEG measures (alpha vs beta) into an
intuitive sensory experience of stress versus relaxation. The project also reinforced principles of
real-time visualization, such as smoothing noisy data, scaling motion by signal intensity, and
designing visual metaphors that match psychological states.
Future Improvements
● Add gradual color blending instead of hard switching between modes.
● Include subtle particle size or opacity changes linked to overall calmness level.
● Experiment with live Muse input to explore individualized thresholds.
● Optionally add sound modulation for a multimodal neurofeedback experience.
Summary
Zen Blood Particles transforms EEG data into a visceral display of the brain’s changing internal state.
By mapping stress to speed and color intensity, and calmness to slow, colorful drift, the animation
becomes a live mirror of the user’s mental tension and relaxation — a minimal yet powerful
demonstration of brain-computer interaction through art.

Sarah Badran/1782750
Course: PSYCH 403A1 — Neuroimaging and Neurostimulation
Part 1: Adaptive Flow Neurofeedback: Beyond Basic Control
The project is based on extending the previous Assignment 2 particle system to a neurofeedback -
based visual output that adjusts and updates in real - time to the simulated EEG input. Using
BrainImation's Attention ( β) and Meditation ( α) variables, I compute engagement = β - α and
determine a flow window near 0.06. The program continuously smooths incoming values to
prevent flickering, as well as calculates distance from the target, and does modifies each session's
difficulty through real - time adaptive tolera nce bands .
When the user drifts out of flow (large β - α) the animation becomes stormy and chaotic and the
center cue reads, “Attention Fluctuating - Recenter.” As values move closer together, the system
enters a calm concentration zone, displays a smooth particle mo vement, and reads, “Focus
Stabilized .” The HUD provides a readout of tolerance metrics, engagement, score, and streak for
current status feedback about their performance and progress (real time neurofeedback).
The design of this model works as a closed - loop BCI system, rather than simple parameter
mapping to visuals in the form of a BCI interface. The animation provides a link between EEG -
derived signals to adaptive visual content output, and quantitative metri cs to train focused
attention engagement to further move the user toward a more stable, even state.

While "Focus Stabilized" can show up when Attention and Meditation have numbers that are not
numerically equal, this is intentional. The program defines a flow zone, where Attention is
slightly higher than Meditation ( β − α ≈ 0.06), to indicate engaged but relaxed mental states,
rather than indicating that those two numbers are truly equal. Because the values are smoothed
over time, the system is responding to trends and averages rather than to single instantaneous
numbers, thereby producing more natural an d stable feedback. This reflects how real
neurofeedback systems reward us when we have good balance with our focus, and when we
achieve sustained mental alertness without overstimulation, rather than focusing on maintaining
absolute stillness.

Sarah Badran/1782750
Course: PSYCH 403A1 — Neuroimaging and Neurostimulation
Part 2
Component & Rationale. I recorded the N400, a negative - going ERP that peaks ~400 ms post -
meaningful stimuli. This component is larger (more negative) for semantically incongruent
versus congruent targets (e.g. CAT – TREE vs. CAT – DOG). The N400 component ind exes
semantic integration difficulty and violation of expectancy.
Stimulus Design & Timing. Each trial consists of prime → target word pairs selected from equal
numbers of match and mismatch items in a balanced design. I utilize a simple state machine with
four defined states: baseline ( – 200 – 0 ms), prime (400 ms), target (400 ms), and ITI (1000 ms).
The timing is compliant with the midterm having a pre - stimulus baseline and post - stimulus
period of ~800 ms in ~1 s epochs at 256 Hz (Muse rate).
At the onset of each target, I sample 256 samples in each of the four channels (TP9, AF7, AF8,
and TP10), run a baseline correction on the 200 – 0 ms, and add the correct epoch to the buffer
allocated to that condition (for example, match/mismatch). I run a moving average for the
specific condition and each channel capped at 20 trials so that the waveform is stable yet
responsive on the display. In the upper part of the display, it shows the condition averages for
each channel, and in the bottom part of the d isplay shows the current prime/target and a label
indicating the condition. Tick marks appear every 100 ms, and I also include an anchor time for 0
ms as a vertical line. I include running trial counters for each condition. Overall Satisfying the
ERP visua lization requirements.

With ~20 – 50 trials/condition, I expect a more negative deflection ~300 – 500 ms post - target for
mismatch versus match, largest over central – parietal regions (Muse coverage is limited but will
show a condition difference). I would expect the baseline to be ne ar 0 μV pre - stimulus; more
trials will help with SNR.