Discussion

This study combined an advanced vascular finger phantom with Monte Carlo modelling to investigate how skin pigmentation alters PPG signals in both reflectance and transmittance measurement modes. Building on prior phantom work that demonstrated a decrease in signal quality with increasing skin pigmentation in reflectance measurements,12 the present design extends the experimental capability to incorporate transmittance mode PPG signals and directly links observed signal changes to simulated photon transport and absorption behavior.

Although pulse oximetry applications have highlighted the potential impact of skin pigmentation in optical measurements, the present study focuses specifically on the underlying PPG signal quality and its dependence on light–tissue interaction. To achieve this, fixed acquisition conditions were employed to isolate the intrinsic optical effects of pigmentation, enabling controlled comparison of wavelength-dependent attenuation and signal quality metrics without confounding effects from adaptive system responses. This approach aims to provide insight into the fundamental mechanisms governing PPG signal formation. This is particularly relevant for emerging wearable technologies that use algorithms based on PPG waveform morphology and derived features, such as blood pressure estimation,40,41 arterial stiffness assessment,42–44 and other cardiovascular markers, where reduced signal quality and increased variability may directly impact algorithm performance.

The multilayer vascular architecture enables a more physiologically representative interaction between light and pulsatile blood volume by incorporating both superficial and deeper vascular layers. Unlike the previous single-vessel phantom, the present design incorporates three distributed perfusion layers spanning superficial and deeper tissue regions, allowing evaluation of wavelength-dependent photon penetration within a more physiologically representative vascular environment. Importantly, the added vascular complexity did not change the expected optical behavior that melanin-simulating absorbers attenuate shorter wavelengths more strongly, as this could also be demonstrated using simpler single-vessel models.8,10 The purpose of the present vascular network was instead to address a limitation of such simplified models: their restricted vascular interaction volume and non-physiological depth distribution of pulsatile blood. This distributed vascular structure increases light-blood interaction across tissue depths, enabling stable acquisition of PPG signals across reflectance and transmittance modes and allowing more consistent comparison of signal quality metrics across wavelengths and skin tones. This includes conditions where signal generation was particularly challenging in our previous study,12 such as shorter wavelengths under dark skin pigmentation, where in vivo signals are still typically observable. This was made possible by the 3D-printed PVA sacrificial template fabrication method, which enables the formation of small, compliant, geometrically complex channel networks that would be difficult to achievable through conventional single-vessel casting approaches. The resulting distributed vascular network more closely reflects the spatial organization of physiological microvascular perfusion within the highly vascularized fingertip, representing a fundamental fabrication advance over prior simplified phantom designs. Therefore, this specific advancement is not the demonstration that melanin absorption is wavelength dependent, which is already well established, but the ability to produce measurable PPG signals within a more physiologically representative depth-distributed vascular environment and to evaluate how distributed pulsatile blood volume interacts with wavelength-dependent photon penetration in both reflectance and transmittance geometries, which could not be reliably achieved using the previous single-vessel configuration.

Across the in vitro measurements, increasing pigmentation produced a consistent reduction in PPG strength, which is expected as higher skin absorption reduces the light available to interrogate pulsatile blood volume and reach the detector, as demonstrated in prior studies.7,45 This effect was most apparent at shorter wavelengths and in reflectance mode, where the detected photons are most influenced by superficial losses, whereas longer wavelengths better preserved pulsatility due to deeper tissue interaction.46 The integrated simulations provide a direct physical insight into the underlying photon transport mechanisms responsible for these trends. By linking experimental observations with photon-based metrics, this work extends previous phantom studies beyond qualitative signal comparison.

Monte Carlo simulations provided a mechanistic explanation for the experimentally observed signal attenuation with increasing skin pigmentation. Fluence distributions demonstrated that higher skin absorption progressively reduced photon energy reaching deeper vascular layers, particularly at shorter wavelengths. This behavior was quantified through the skin contribution index (SCI), which increased with pigmentation and decreased with wavelength. These findings are consistent with prior modeling studies of light–tissue interaction in PPG, which show that melanin-dominated absorption redistributes photon energy toward superficial layers, thereby limiting sensitivity to pulsatile blood volume changes.17,47

Similar trends have also been reported in Monte Carlo studies of pulse oximetry, where increasing melanin concentration leads to a reduction in AC/DC amplitude and a systematic drift in the ratio-of-ratios, driven by wavelength-dependent attenuation of the pulsatile signal.48 In particular, these studies demonstrate that higher melanin levels reduce the detected pulsatile component and alter the relative contribution of red and infrared wavelengths, contributing to a bias in SpO2 estimation. The present results are in agreement with these observations, as the experimentally measured reductions in PPG amplitude, SNR, and AC/DC ratio with increasing pigmentation reflect the same underlying mechanism of reduced photon penetration and diminished interaction with pulsatile blood volume.

The 50% cumulative absorption depth increased with wavelength, indicating progressively deeper photon interaction at red and infrared wavelengths. Across all wavelengths, darker pigmentation resulted in slightly shallower penetration depths, reflecting increased superficial absorption. Energy balance analysis showed that absorption was the dominant mechanism of photon fate at all wavelengths and pigmentation conditions. Together, these results demonstrate that pigmentation primarily affects PPG signal formation by altering where photon energy is deposited within the tissue.

Strong correlation between Monte Carlo predictions and experimental PPG metrics validates the combined modeling-phantom approach. Simulated diffuse reflectance and transmittance closely matched measured PPG amplitudes, while SCI showed strong agreement with AC/DC ratios, particularly in reflectance mode. Although the correlation analysis is descriptive in nature, reflecting the three discrete and physically distinct skin pigmentation levels from the experimental design, the consistent directional agreement between Monte Carlo predictions and experimental PPG metrics across multiple wavelengths and measurement modes supports the validity of the combined modelling-phantom approach. Notably, cases where a correlation was not expected, such as the 530 nm transmittance condition where signal penetration is limited, did not exhibit meaningful alignment, further indicating that the reported Pearson correlations reflect genuine optical behavior rather than coincidental agreement. These findings demonstrate that skin pigmentation-driven changes in photon transport are directly reflected in measured PPG features, strengthening the confidence that the observed attenuation trends arise from optical mechanisms rather than artifacts of phantom design or experimental variability. Such agreement is rarely demonstrated in phantom studies and highlights the value of integrating modeling alongside in vitro experimentation.

One limitation of this study is the simplified blood optical properties, as the blood-mimicking fluid exhibits a monotonic absorption spectrum, rather than the wavelength-dependent hemoglobin absorption spectra observed in vivo,49 which vary with oxygenation level, and does not replicate the high scattering coefficient of red blood cells. However, this blood-mimicking fluid preserves the expected trend of higher absorption at shorter wavelengths and lower absorption in the infrared region, allowing relative comparisons of signal behavior. Similar nigrosine-based blood-mimicking fluids have been used in optical phantom studies to investigate light–tissue interactions in in vitro systems.8,10 Regarding scattering, while red blood cells contribute substantially to blood scattering in vivo, this effect predominantly governs photon interaction within the vascular compartment; consequently, increased blood scattering would not be expected to alter the directional relationship between skin pigmentation and PPG signal attenuation reported in this study. Nevertheless, incorporating physiologically representative blood optical properties, including wavelength-dependent hemoglobin absorption and realistic red blood cells scattering, remains an important goal for future phantom improvements, particularly for extending this framework toward pulse oximetry applications.

A second limitation relates to the use of discrete wavelengths in the Monte Carlo simulations, whereas the experimental PPG system employs LED sources with finite spectral bandwidths. In LED-based PPG systems, the measured signal represents the convolution of the LED emission spectrum, tissue optical properties, and detector sensitivity, which may lead to a spectral shift in darker skin tones, particularly in the red spectral region.50 As a result, the simulations are more representative of narrowband illumination and do not fully capture the spectral response of the experimental setup.

A third limitation is that the scattering properties of the skin and adipose tissue layers do not fully reproduce the wavelength-dependent decrease in scattering typically observed in biological tissue. The greatest divergence from reported literature values was observed at 530 nm, where TiO2 particle scattering characteristics differ from the scattering behavior of biological tissue constituents such as collagen fibers and cell organelles. At this wavelength, an underestimation of scattering would reduce photon pathlength in the superficial layer, potentially leading to an underestimation of the filtering effect of skin pigmentation on the detected signal. However, as melanin absorption rather than scattering is the dominant driver of superficial attenuation at 530 nm, the directional trends in skin pigmentation-driven signal attenuation reported here are not expected to be fundamentally altered. Future work will focus on incorporating wavelength-dependent scattering behavior and physiologically representative oxygenated and deoxygenated blood levels to further improve optical realism and extend the framework.

Overall, this study contributes to the phantom-based literature8,10,11,13,51,52 by providing a quantitative and mechanistic assessment of how skin pigmentation influences PPG signal quality. Although previous phantom studies have incorporated varying skin tone layers, the present work extends this approach by evaluating specific PPG features, including peak-to-peak amplitude, SNR, and AC/DC ratio, under controlled and physiologically relevant perfusion conditions. By integrating the advanced vascular phantom with Monte Carlo modelling, this study offers a robust framework to isolate the optical effects of skin pigmentation on PPG quality.