J. Eur. Opt. Society-Rapid Publ. 22, 48( 2026) 471
systematically examined across different photon transport regimes or for physiologically realistic pore geometries.
The present study examines the influence of skin pore microgeometry on NIR photon transport using anatomically realistic, MRI – segmented cadaveric head models and complementary simulation frameworks. Photon transport in biological tissue exhibits a transition from an initial quasi-ballistic regime near sources and interfaces to a diffusive regime as multiple scattering events accumulate, a behavior well established in radiative transfer theory and Monte Carlo( MC) modeling of turbid media [ 21, 22 ]. Accordingly, photon propagation across the source – scalp interface and subsequent transport within tissue were modeled using a cascaded simulation framework. Ballistic photon propagation simulations with explicit representation of physiologically relevant pore geometries were first used to resolve surface-scale optical interactions and determine the resulting photon states. These photon states, characterized by their position, direction, and weight, were then used to initialize MC-based diffusive photon transport simulations within anatomically realistic head models. By linking surface-scale geometric optics with subsequent scatteringdominated transport in deeper tissues, this approach enables systematic assessment of how pore-scale surface features influence NIR photon transport and clarifies the conditions under which such features may be neglected or require explicit representation in optical tissue imaging models.
2 Material and methods
The study was conducted using a cascaded simulation workflow, consisting of subsections: Magnetic Resonance Imaging – Derived Cadaveric Head Anatomy( Sect. 2.1), Ballistic Photon Propagation Simulations( Sect. 2.2), Diffusive Photon Transport Simulations( Sect. 2.3), Data Processing and Quantitative Analysis( Sect. 2.4), as detailed in the following subsections.
2.1 Magnetic resonance imaging – derived cadaveric head anatomy
Anatomically realistic cadaveric head geometries used in this study were obtained from a previously published MRI dataset originally acquired for the study DrSVision: A Machine Learning Tool for Cortical Region – Specific fNIRS Calibration Based on Cadaveric Head MRI [ 23 ]. The dataset comprises eight adult human cadaveric heads( five male and three female; age range: 67 – 97 years), selected to exclude cranial trauma, prior surgery, or visible pathology. Cadaveric material was obtained through a licensed anatomical supplier with informed consent provided during donors’ lifetimes, and all procedures were conducted in accordance with ethical and legal guidelines, as detailed in the original study.
High – resolution T1 and T2 – weighted 3 Tesla MRI scans enabled segmentation of scalp, skull, cerebrospinal fluid( CSF), and brain tissues relevant to optical modeling. In the present work, the previously segmented geometries and tissue layer thicknesses were reused directly for diffusive photon transport simulations. No new MRI acquisitions or anatomical segmentation procedures were performed.
2.2 Ballistic photon propagation simulations
Ballistic photon propagation at the skin surface was modeled using the COMSOL Multiphysics( v6.2, COMSOL AB, Stockholm, Sweden) [ 24 ] platform, employing the geometric optics interface in 2-dimensional( 2D) space with explicit representation of physiologically relevant pore geometries with surface roughness to resolve surface-scale optical interactions and determine the resulting photon states. These simulations were performed using representative surface geometries, as the underlying anatomical layers are not involved at this stage. The geometric optics interface operates in the high – frequency limit where the vacuum wavelength is negligible compared to the characteristic dimensions of the geometry. In this regime, Maxwell’ sequations are reduced to the Eikonal equation [ 25 ],
jrSj 2 ¼ n 2 ðÞ r; ð1Þ
where S represents the optical phase and n( r) is the local refractive index. This formulation treats photon propagation as a set of discrete rays whose trajectories are determined by integrating the Hamiltonian form of the ray equations, dq dt ¼ @ x
@ k; dk dt ¼� @ x
@ q; ð2Þ ð3Þ
where q, k, and t denote the position vector, wave vector, and time of flight, respectively. The term x signifies the angular frequency of the radiation, which acts as the Hamiltonian of the system and remains constant along the ray path. To account for energy transport and the formation of caustics within the absorbing media, the local intensity I was computed by tracking the evolution of the principal radii of curvature( r 1, r 2) of the wavefront. For a medium characterized by a complex refractive index, the intensity I along the ray path is governed by,
Z r 1; 0 r 2; 0 I ¼ I 0 r 1 r exp � l a ds; ð4Þ
2
where I 0, r 1, 0, and r 2, 0 are the initial intensity and principal radii of curvature, respectively. The exponential term accounts for attenuation via the Beer – Lambert law, where the absorption coefficient l a is integrated along the incremental path length ds. This curvature – based approach ensures a mesh – independent evaluation of the energy flux, even in regions of high convergence where traditional ray – counting methods fail.
A scalp tissue layer was represented using a rectangular geometry with a thickness of 5.64 mm, corresponding to that of the mean cadaveric head( Supplementary Materials,