← Back to Week 10 · All Course Demos

fNIRS Photon Transport & Short-Channel Scalp Regression

Diffuse Optical Tomography · Monte Carlo Photon Random Walks, Sensitivity Banana Paths & Superficial Physiological Decontamination

The Scalp Contamination Problem: Near-infrared light diffuses through layered biological media in a characteristic curved "banana path". However, standard long-separation channels (\(d \approx 30\text{ mm}\)) pass twice through superficial extracerebral tissue (scalp and skull). Up to 80% of recorded hemodynamic oscillations arise from systemic scalp Mayer waves (\(\approx 0.1\text{ Hz}\)) rather than cortical brain activity. Modern fNIRS solves this by incorporating short-separation channels (\(d \approx 8\text{ mm}\)) to measure and regress out superficial contamination.

12.2 mm

Mean Penetration · Cortex

5.85

Diff. Pathlength (DPF)

+18.4 dB

Cortical Signal Quality

Tissue Slab & Photon Transport Simulation

Source (Emitter)

Short Optode (8 mm)

Long Optode (30 mm)

Photon Sensitivity (Banana Path)

Real-Time Hemodynamic Signal Decontamination

Short Channel (Scalp Mayer Wave)

Raw Long Channel (Corrupted Mix)

Decontaminated Cortical Signal

Task Paradigm Mode

Long Optode Separation (d_SD) 30 mm

15 mm (Superficial) Cortex Reached 45 mm (Deep)

Tissue Scattering (μ'_s) 1.0 mm^-1

0.5 mm⁻¹ (Low / Ballistic) 2.5 mm⁻¹ (Turbid / Diffuse)

Cortical Neural Activation (ΔHbO₂) 1.8 μM

0.0 (Resting State) 3.0 (Robust Activation)

Scalp Mayer Wave Amplitude High

0.0 (Clean Scalp) 2.5 (Severe Noise)

Enable Short-Channel Regression

Clinical & Biophysical Insight:

Notice how the Short Optode (8 mm) samples only the yellow scalp layer. Try unchecking "Enable Short-Channel Regression" with High Mayer waves to observe how blood pressure oscillations masquerade as cortical activation. Vary Tissue Scattering to observe how photon wander spreads the banana path!

Diffuse Optical Physics & Signal Decontamination

1. Monte Carlo Radiative Transport:

In biological tissue, the transport mean free path (\(1/\mu_s' \approx 1\text{ mm}\)) is orders of magnitude shorter than the absorption length (\(1/\mu_a \approx 100\text{ mm}\)). Photons undergo thousands of elastic scattering collisions before being detected or absorbed. The spatial probability distribution of detected photons follows the diffusion equation, producing the characteristic curved "banana path" whose mean penetration depth scales approximately as \(z_{\text{mean}} \approx \frac{1}{2} \sqrt{d_{\text{SD}}}\).

2. The Extracerebral Contamination Hazard:

Because photons enter and exit through the head's surface, they traverse the highly vascularized scalp and skull twice. Scalp blood vessels exhibit pronounced systemic physiological oscillations: Mayer waves in blood pressure (\(\sim 0.1\text{ Hz}\)), respiration (\(\sim 0.25\text{ Hz}\)), and cardiac pulsatility (\(\sim 1.0\text{ Hz}\)). Because psychological stress, speech, and motor effort modulate autonomic blood pressure, these scalp oscillations correlate with task blocks, producing catastrophic false-positive activation maps.

3. Short-Separation Optode Physics:

By placing a detector at \(d_{\text{SD}} \le 8\text{ mm}\) from the emitter, the sensitivity profile is strictly confined within the \(3\text{--}4\text{ mm}\) scalp depth, entirely bypassing the underlying cerebral cortex. This provides a pure, real-time reference recording of extracerebral systemic physiology without any contamination from true brain activity.

4. Adaptive GLM Short-Channel Filtering:

To eliminate superficial noise, the long-channel optical density is modeled via multilinear regression:

Y_{\text{long}}(t) = \beta_{\text{task}} X_{\text{task}}(t) + \beta_{\text{scalp}} Y_{\text{short}}(t) + \epsilon(t)

Regressing out \(Y_{\text{short}}(t)\) strips away the confounding systemic Mayer waves and respiratory noise, isolating the true cortical hemodynamic response.