CLNF - Closed-Loop Neurofeedback¶

CLNF uses real-time calcium imaging (ΔF/F) to provide audio feedback correlated with neural activity in a target brain region.

How It Works¶

Camera → ΔF/F Calculation → Threshold Check → Audio Feedback + Reward

1. Imaging: Wide-field calcium imaging of cortical activity
2. Processing: Calculate ΔF/F (delta F over F) in real-time
3. Feedback: Map activity level to audio tone frequency
4. Reward: Deliver water when activity exceeds threshold

Running CLNF Experiments¶

Single ROI Experiment¶

python brain/cla_reward_punish_1roi.py

Monitors a single brain region. Reward triggered when activity exceeds threshold.

Dual ROI Experiment¶

python brain/cla_reward_punish_2roi.py

Monitors two brain regions (can be extended to more number of regions) with mathematical operations such as:

• Addition: ROI1 + ROI2
• Subtraction: ROI1 - ROI2

Other operations such as multiplication, division etc. are supported and can be specified in the config.ini file.

Configuration Parameters¶

Required Settings (config.ini)¶

[raspicambrain_reward_punish_1roi]
vid_source = PiCameraStream
data_root = /path/to/data/
resolution = 256, 256
framerate = 15
dff_history = 6
ppmm = 25.6
bregma = 135, 128
roi = M1_L
roi_size = 0.3, 0.3
audio = 1
n_tones = 18
reward_threshold = 0.075
adaptive_threshold = 1
total_trials = 60
max_trial_dur = 30
success_rest_dur = 10
fail_rest_dur = 15
initial_rest_dur = 30

Parameter Reference¶

Available Brain Regions¶

Pre-configured seed locations (in mm from bregma):

seeds_mm = {
    "ALM": {"ML": 1.5, "AP": 2.5},
    "M1": {"ML": 1.8603, "AP": 0.64181},
    "M2": {"ML": 0.87002, "AP": 1.4205},
    "FL": {"ML": 2.4526, "AP": -0.5668},
    "HL": {"ML": 1.6942, "AP": -1.1457},
    "BC": {"ML": 3.4569, "AP": -1.727},
    "V1": {"ML": 2.5168, "AP": -3.7678},
    "RS": {"ML": 0.62043, "AP": -2.8858},
    # ... and more
}

Audio Feedback Mapping¶

Neural activity is mapped to audio frequency:

ΔF/F = 0    →  1000 Hz (low tone)
ΔF/F = threshold → 1000 × 2^(n_tones/4) Hz (high tone)

The system uses quarter-octave frequency increments from 1-24 kHz.

Session Types¶

Configure reward behavior in code:

# Normal audio + normal reward
sessionType = clh.SessionType.normal_audio_normal_reward

# No audio + random reward (control)
sessionType = clh.SessionType.no_audio_random_reward

Adaptive Thresholding¶

When adaptive_threshold = 1:

• Every 30 seconds, check rewards in epoch
• If > 1 reward: increase threshold by 0.002
• If 0 rewards: decrease threshold by 0.002

This maintains consistent reward rates across sessions.

Hardware Setup¶

Required Components¶

• Raspberry Pi 4B+
• Raspberry Pi Camera (v2)
• MPR121 capacitive touch sensor (lick detection)
• LED array (behavior, reward, fail indicators)
• Water reward valve
• Audio speaker

GPIO Pinout¶

I2C Connections¶

• MPR121 on I2C bus (SCL/SDA)

Data Output¶

VideoTimestamp.txt Format¶

frame   time    roi_dff freq    rew_threshold   reward  trial   audio   lick

Summary CSV¶

mouse_id, session, data_path, start, end, duration, fps, dff_history, n_tones, reward_threshold, rewards, audio, session_type

Example Workflow¶

1. Edit config.ini with your parameters
2. Run: python brain/cla_reward_punish_1roi.py
3. Enter mouse ID when prompted
4. Adjust ROI positions in preview window
5. Press Esc to start session
6. Monitor real-time feedback
7. Press Esc to end or wait for completion

Troubleshooting¶

See Troubleshooting Guide for common issues.