edVirtus Reference

What Are Computational RF Dosimetry, FDTD, and FEM?

How Do Numerical Models Calculate Internal RF Fields and Absorption?

Computational RF dosimetry uses numerical solutions of Maxwell's equations to estimate fields, currents, absorbed power, and absorbed energy in or at the human body. It is valuable where internal quantities cannot be measured directly in a person or where many configurations must be compared.

A model combines a source, surrounding geometry, numerical domain, and representation of the body. The source must reproduce the relevant frequency, power, modulation or time behavior, polarization, antenna currents, and position. Environmental objects may be included when they materially change the field.

The finite-difference time-domain method, or FDTD, divides space into a three-dimensional grid and advances electric and magnetic fields in small time steps. One simulation can contain broadband information, and the method works well with detailed voxel models derived from medical images.

FDTD accuracy depends on cell size, time step, material assignment, source representation, and treatment of the outer boundary. Fine cells resolve small anatomical or source features but increase memory and run time. Staircasing of curved surfaces and poorly resolved tissues can introduce error.

The finite-element method, or FEM, divides the modeled region into interconnected elements and solves for the fields using basis functions. Flexible, locally refined meshes can represent curved boundaries, thin structures, antennas, waveguides, and complex equipment efficiently.

FEM is often applied in the frequency domain, making it well suited to individual frequencies and resonant structures. A wide frequency sweep may require repeated solutions. Accuracy depends on mesh quality, element order, material properties, boundary conditions, and solver convergence.

The Method of Moments, or MoM, converts integral equations for currents or charges on conducting or equivalent surfaces into a matrix problem. It can model antennas and scattering structures efficiently when the important behavior is concentrated on surfaces, but detailed heterogeneous anatomy is usually better suited to a volume method such as FDTD or FEM.

Ray-based methods approximate propagation as rays that reflect, diffract, transmit, and scatter. Ray tracing can be useful for large environments at frequencies for which the objects are electrically large, such as rooms, vehicles, platforms, and urban spaces. It is generally unsuitable for calculating detailed internal absorption by itself, particularly in reactive near fields or where full-wave coupling is important.

Hybrid methods combine techniques at interfaces where each is most effective. A site-scale ray model might estimate the field incident on a person, while FDTD or FEM calculates coupling and absorption within a detailed computational human model. Surface and volume methods can also be coupled to represent an antenna, surrounding structures, and anatomy without applying one computational method inefficiently to the entire problem.

No method is universally superior. FDTD may suit broadband pulses and voxel anatomy; FEM may suit irregular geometry and local refinement; MoM may suit antennas and conducting surfaces; and ray-based methods may suit electrically large environments. Method selection should follow the assessment objective, scale, frequency range, field region, available validation, and required dosimetric output.

Computational human models range from simple homogeneous shapes and standardized phantoms to anatomically detailed, tissue-specific voxel or surface representations. Tissue dielectric properties must correspond to frequency, and posture, body size, age, and source position should represent the assessed population or a justified conservative case. Greater anatomical detail is useful only when it is supported by suitable material data, numerical resolution, and validation.

Typical outputs include induced electric field, whole-body or localized specific absorption rate, specific energy absorption, absorbed power density, and absorbed energy density. Results must be processed using the spatial and temporal averaging rules specified for the applicable basic restriction.

Verification asks whether the equations were solved correctly. It includes mesh refinement, time-step or solver convergence, energy balance, boundary checks, and comparison with analytical solutions or benchmark problems. A visually plausible field plot is not sufficient evidence of numerical reliability.

Validation asks whether the model represents physical reality adequately for its intended use. Comparison may use phantom measurements, calibrated field measurements, published interlaboratory studies, or standardized validation procedures. Agreement should be judged together with measurement and model uncertainty.

Uncertainty can arise from tissue properties, anatomy, positioning, source power and geometry, mesh resolution, numerical settings, and post-processing. Sensitivity studies help identify dominant contributions and distinguish genuine physical variation from numerical artifacts.

A reproducible report identifies the software and version, numerical method, geometry, mesh, material data, source and operating state, boundaries, convergence tests, validation evidence, averaging, uncertainty, and any conservative assumptions. Input files and scripts should be controlled records where appropriate.

Computational dosimetry complements rather than automatically replaces measurement. Measurements can confirm source output and external fields, while computation reveals internal distributions that probes cannot access. A combined approach often provides the strongest assessment.

The purpose is not to produce the most elaborate model possible. It is to use a verified, validated, and appropriately conservative model that answers the safety question with sufficient accuracy and transparent limitations.