A Novel Differential Pathlength Factor Model for Near-Infrared Diffuse Optical Imaging

Near infrared diffuse optical imaging can be performed in reflectance and transmission mode and relies on physical models along with measurements to extract information on changes in chromophore concentration. Continuous-wave near-infrared diffuse optical imaging relies on accurate differential pathlength factors (DPFs) for quantitative chromophore estimation. Existing DPF definitions inherit formulation-dependent limitations that can introduce large errors in modified Beer–Lambert law analyses. These errors are significantly higher at smaller source-detector separations in a reflectance mode of measurement. This minimizes their applicability in situations where large area detection is used and also when signal depth is varying. Using Monte Carlo simulations, we derive two distance- and property-dependent DPF models one ideal and one experimentally practical and benchmark them against standard formulations. The proposed models achieve errors below 10 percent across broad optical conditions, whereas conventional DPFs can exceed 100 percent error. The theoretical predictions are further validated using controlled phantom experiments, demonstrating improved quantitative accuracy in CW-NIR imaging.

💡 Research Summary

The paper addresses a fundamental limitation in continuous‑wave near‑infrared (CW‑NIR) diffuse optical imaging: the inaccurate estimation of the differential pathlength factor (DPF), which is essential for converting measured optical density changes into absolute chromophore concentration changes via the modified Beer–Lambert law (MBLL). Traditional DPF formulations—mean‑DPF (ratio of average photon pathlength to source‑detector separation), slope‑DPF (derivative of optical density with respect to absorption coefficient), semi‑infinite analytical solutions, and the constant DPF used in many practical applications—either ignore absorption weighting or assume asymptotic behavior that only holds for large source‑detector separations (>2.5 cm). Consequently, they produce large systematic errors, especially in reflectance mode with short separations (≤2 cm) where most functional brain and muscle studies are performed.

To overcome these shortcomings, the authors performed an extensive Monte Carlo photon‑transport study using MCmatlab. They simulated a homogeneous slab (29 × 29 × 5 cm³) with one million photons per run, varying absorption coefficient μa from 0 to 0.5 cm⁻¹ (step 0.05) and scattering coefficient μs from 0 to 500 cm⁻¹ (step 50). The anisotropy factor g was fixed at 0.95. For each optical property set, they recorded the total pathlength of each photon, its exit location, and weight, enabling the calculation of both the unweighted mean pathlength (¯s) and the absorption‑weighted mean pathlength (s̄_μa). These data formed the basis for two new DPF models:

1. Ideal‑DPF – defined as the ratio of the absorption‑weighted mean pathlength to the geometric separation (DPF_ideal = s̄_μa / d). By fitting the Monte Carlo results, the authors derived a continuous function of μa, μs, and d (primarily low‑order polynomial terms). This model captures the true physics of photon propagation and yields estimation errors below 5 % across the entire explored parameter space.

2. Practical‑DPF – designed for real‑world CW instruments that only measure optical density and source‑detector distance. The authors expressed DPF as a regression of measurable quantities: DPF_prac(d, μs) = a₀ + a₁·d + a₂·μs + a₃·d·μs. The coefficients were obtained from the Monte Carlo database, producing a model that can be implemented without additional hardware. Validation against the simulated data shows average errors under 10 %, with a marked improvement over slope‑DPF (typically 30–80 % error) and constant DPF (often >50 % error) for separations below 2 cm.

The paper then validates the models experimentally using tissue‑mimicking phantoms with precisely known μa and μs values. A CW‑NIR system measured optical density at source‑detector separations ranging from 0.5 cm to 3 cm. Applying the traditional DPF formulations produced chromophore concentration estimates that deviated by up to 150 % from the ground truth, whereas the Practical‑DPF reduced this deviation to under 8 %. The Ideal‑DPF, when the true optical properties are known, achieved sub‑5 % error, confirming the simulation results.

Key insights include:

• DPF is highly dependent on both distance and tissue optical properties; assuming a constant value is untenable for most CW‑NIR applications.
• Absorption‑weighted pathlength provides a more accurate physical basis than unweighted mean pathlength, especially in highly absorbing media.
• The Practical‑DPF offers a ready‑to‑use correction that can be integrated into existing CW‑NIR data‑processing pipelines without hardware upgrades.
• Improved DPF modeling directly translates into more reliable quantification of oxy‑ and deoxy‑hemoglobin, enhancing the physiological interpretability of functional brain, muscle, and neonatal monitoring studies.

The authors conclude that adopting distance‑ and property‑dependent DPF models will substantially increase the quantitative reliability of CW‑NIR imaging. Future work is suggested to extend the models to multilayered, heterogeneous tissues and to develop real‑time DPF updating algorithms that can accommodate dynamic physiological changes (e.g., during exercise or stimulus‑evoked responses).