DUNEuro -- A software toolbox for forward modeling in bioelectromagnetism

Accurate and efficient source analysis in electro- and magnetoencephalography using sophisticated realistic head geometries requires advanced numerical approaches. This paper presents DUNEuro, a free and open source C++ software toolbox for forward modeling in bioelectromagnetism. Building upon the DUNE framework, it provides implementations of modern fitted and unfitted finite element methods to efficiently solve the forward problems in electro- and magnetoencephalography. The user can choose between a variety of different source models that are implemented. The software’s aim is to provide interfaces that are extendible and easy-to-use. In order to enable a closer integration into existing analysis pipelines, interfaces to Python and Matlab are provided. The practical use is demonstrated by a source analysis example of somatosensory evoked potentials using a realistic six compartment head model.

💡 Research Summary

**
The paper introduces DUNEuro, an open‑source C++ toolbox built on the DUNE (Distributed and Unified Numerics Environment) framework for forward modeling in bioelectromagnetism, specifically EEG and MEG. Accurate source analysis in EEG/MEG requires realistic head models with multiple tissue compartments and sophisticated numerical methods. Existing open‑source packages either provide only low‑order finite element (FE) implementations or lack flexibility for extending discretization schemes. DUNEuro addresses these gaps by offering both fitted (geometry‑conforming) and unfitted (geometry‑independent) finite element methods, a variety of source‑model formulations, and high‑level interfaces for Python and Matlab.

The toolbox distinguishes two major categories of discretization. The fitted category uses meshes that exactly match tissue boundaries and currently implements continuous Galerkin (CG‑FEM) and discontinuous Galerkin (DG‑FEM) schemes. The unfitted category employs level‑set representations of tissue interfaces, allowing the mesh to be independent of the anatomy; it currently provides an unfitted DG (UDG) method based on CutFEM ideas. Both categories share a common driver interface (MEEGDriverInterface) that abstracts the solution process. Specific drivers (FittedMEEGDriver, UnfittedMEEGDriver) are instantiated with two template parameters: a Solver (handling matrix assembly and linear system solution) and a SourceModelFactory (building the right‑hand side for a chosen source model). This design enables developers to add new solvers (e.g., multigrid, direct sparse) or new source formulations (partial‑integration, St Venant, subtraction, Whitney) with minimal code changes.

A key computational concept is the transfer matrix. Since EEG/MEG forward problems only require potentials or magnetic fields at sensor locations, DUNEuro computes a matrix T = R A⁻¹ (where A is the system matrix and R extracts sensor values). Once T is available, any number of source configurations can be evaluated by a simple matrix‑vector multiplication, dramatically reducing runtime for large source spaces.

The implementation reuses several mature DUNE modules: dune‑uggrid for conforming tetrahedral/hexahedral meshes, dune‑subgrid for extracting sub‑meshes from voxel images, dune‑pdelab for generic discretizations, dune‑udg for unfitted DG, and dune‑istl for iterative solvers. This modularity keeps memory consumption low and leverages DUNE’s extensive testing infrastructure.

To demonstrate practical utility, the authors apply DUNEuro to a realistic six‑compartment head model (skin, skull, CSF, gray matter, white matter, and brain‑stem) and analyze somatosensory evoked potentials. They compute forward solutions, construct the EEG transfer matrix, and perform source localization using a minimum‑norm estimator. Results are compared with established pipelines (FieldTrip, SimBio), showing comparable or improved accuracy while offering faster computation thanks to the transfer‑matrix approach. The Python and Matlab bindings allow the entire workflow to be scripted and integrated into existing analysis pipelines.

In summary, DUNEuro provides a flexible, extensible, and high‑performance platform for forward modeling in EEG and MEG. Its dual support for fitted and unfitted FEM, comprehensive source‑model library, and language bindings make it suitable for researchers and clinicians who need realistic head modeling without deep expertise in C++ FEM programming. Future work may extend the toolbox to other bioelectromagnetic applications such as transcranial electrical or magnetic stimulation, real‑time brain‑computer interfaces, and cross‑platform deployment.