Check detector mesh convergence#

Study how refining a tetrahedral mesh changes detector powers for a layered sample at one wavelength. This is a numerical check to perform before an inverse fit; it does not establish the accuracy of the P1 model for skin.

import os
import numpy as np
import MomentRT

ureg = MomentRT.ureg

Configure the sample and instrument#

The layer thickness and optical properties are illustrative. The epidermis preset’s melanin fraction is a model parameter, not a clinical measurement. Both materials have the same refractive index because the solver does not model internal refractive-index jumps.

wavelength = 650 * ureg.nm

background = MomentRT.Material(
    absorption=0.025 / ureg.mm,
    reduced_scattering=1.2 / ureg.mm,
    refractive_index=1.4,
    phase_function=MomentRT.Isotropic(),
)

epidermis = MomentRT.Epidermis(
    melanin_fraction=0.01,
    wavelength=wavelength,
    refractive_index=1.4,
)

sample = MomentRT.StructuredMedium(
    origin=(-3, -3, 0) * ureg.mm,
    size=(6, 6, 3) * ureg.mm,
)

sample.add_background(material=background)

sample.add_layer(name="Epidermis", thickness=0.5 * ureg.mm, material=epidermis)

source = MomentRT.SurfaceSource(
    face="z_min",
    position=np.array([-1.5, 0, 0]) * ureg.mm,
    radius=0.8 * ureg.mm,
    wavelength=wavelength,
    optical_power=1 * ureg.mW,
)

detectors = {
    name: MomentRT.SurfaceDetector(
        face="z_min",
        position=np.array([position, 0, 0]) * ureg.mm,
        radius=0.75 * ureg.mm,
        collection_axis=[0, 0, -1],
        acceptance_angle=90 * ureg.degree,
    )
    for name, position in (("near", 0.5), ("far", 1.5))
}

boundary = MomentRT.RobinBoundary(effective_reflectance=0.4)

solver = MomentRT.P1(
    boundaries={face: boundary for face in MomentRT.boundary.FACES},
)

mesh = MomentRT.TetrahedralMesh.from_sample(
    sample=sample,
    resolution=0.8 * ureg.mm,
)

simulation = MomentRT.Simulation(
    medium=sample,
    source=source,
    mesh=mesh,
    solver=solver,
    detectors=detectors,
)

Compare successively refined meshes#

The initial mesh belongs to the simulation; the study creates independent meshes at the explicit target resolutions below. It compares detector powers and sampled source and aperture areas. Two consecutive refinements must meet the tolerances at the fine end, and negative outputs remain visible. The plot shades the power tolerance around the finest detector values and area tolerances around the geometric footprint areas. These bands are not sufficient on their own: the study also checks adjacent refinements. The source power is incoming diffuse light already coupled into the sample; detector power uses the modeled Lambertian angular approximation. The progress table prints each completed mesh level as the study runs.

study = MomentRT.MeshConvergence(
    simulation=simulation,
    mesh_resolutions=[
        0.8 * ureg.mm,
        0.4 * ureg.mm,
        0.3 * ureg.mm,
        0.2 * ureg.mm,
        0.15 * ureg.mm,
        0.10 * ureg.mm,
    ],
    power_rtol=0.05,
    area_rtol=0.05,
    required_consecutive=2,
    progress=True,
)

result = study.run()

convergence_figure = result.plot(
    power_unit="uW",
    area_unit="mm^2",
    show=False,
)
Detector mesh convergence at 650 nm
         Mesh      Nodes   Time (s)  Source err        near µW    Δ power   Area err         far µW    Δ power   Area err
-------------------------------------------------------------------------------------------------------------------------
       0.8 mm        479       0.06      0.93%          6.284          —     2.18%         0.9677          —     6.79%
  fluence: Negative nodal fluence.
       0.4 mm      2,086       0.24      0.80%          5.149     22.05%     1.98%         0.8497     13.88%     1.39%
       0.3 mm      4,265       0.52      1.07%          4.791      7.47%     1.72%         0.9222      7.86%     0.31%
       0.2 mm     12,462       2.18      0.76%          4.319     10.93%     1.66%         0.9363      1.50%     1.62%
      0.15 mm     27,861       9.37      0.31%          4.317      0.03%     0.03%         0.9669      3.16%     0.03%
       0.1 mm     87,318     115.28      0.36%          4.398      1.83%     0.11%         0.9851      1.86%     0.38%
Converged: True
Consecutive passing refinements at the fine end: 2 of 2 required.
Tolerances: power 5.00% relative + 0 µW absolute; footprint area 5.00% relative.
Recommended mesh: 0.1 mm target resolution.
Total time: 127.65 s

Inspect the finest solution#

Only recommend a mesh when the final sequence meets the stated criteria. The fluence volume below is interpolated onto a separate display grid and is not a convergence test. Transparent layer surfaces show construction geometry, not a material map. A convergence result at this wavelength and baseline does not establish convergence for other wavelengths or tissue properties, or P1 accuracy.

finest = result.results[-1]

if finest is None:
    raise RuntimeError("The finest source footprint was not resolved.")

background_layer = MomentRT.Layer(
    name="Underlying background",
    depth=0.5 * ureg.mm,
    thickness=2.5 * ureg.mm,
    material=background,
)

gallery_system_figure = finest.fluence.plot(
    view="volume",
    scale="linear",
    display_resolution=0.5 * ureg.mm,
    overlays=[*sample.structures, background_layer, source, *detectors.values()],
    show=False,
)

Show the figures when running locally#

Sphinx-Gallery embeds both figures without opening browser windows.

if os.environ.get("MOMENTRT_SPHINX_GALLERY") != "1":
    convergence_figure.show()
    gallery_system_figure.show()

Total running time of the script: (2 minutes 12.511 seconds)

Gallery generated by Sphinx-Gallery