Accuracy and run-time comparison for different potential approaches and iterative solvers in finite element method...

Area of Science:

• Neuroscience
• Computational Biology
• Medical Imaging

Background:

• Accurate and efficient computational modeling is crucial for medical diagnostics and neuroscience research.
• Electroencephalography (EEG) source reconstruction involves solving the complex EEG inverse problem to map brain activity noninvasively.
• The EEG forward problem simulates head surface potentials from known cortical current sources, often using numerical methods like the finite element (FE) method.

Purpose of the Study:

• To compare the efficiency of Algebraic Multigrid (AMG), Incomplete Cholesky (IC), and Jacobi preconditioners when used with the Conjugate Gradient (CG) method for solving the FE-based EEG forward problem.
• To examine the performance of these solvers in conjunction with different dipole singularity treatment methods (full subtraction, Venant, partial integration).
• To evaluate computational speed versus numerical accuracy in a realistic four-compartment sphere model with an anisotropic skull.

Main Methods:

• Iterative solution of the finite element (FE) method-based EEG forward problem using the Conjugate Gradient (CG) method.
• Comparison of Algebraic Multigrid (AMG), Incomplete Cholesky (IC), and Jacobi preconditioners.
• Evaluation of full subtraction, Venant, and partial integration methods for handling dipole singularities.
• Utilizing specifically tuned constrained Delaunay tetrahedralization (CDT) FE meshes within a four-compartment sphere model.

Main Results:

• The Algebraic Multigrid (AMG)-preconditioned Conjugate Gradient (CG) method demonstrated an order of magnitude increase in computational speed compared to standard preconditioners (IC, Jacobi).
• The computational speed advantage of AMG-CG increased as the mesh size decreased.
• High accuracies were achieved for both full subtraction and direct potential approaches with CDT FE meshes.
• The full subtraction approach yielded the best accuracies when the homogeneity condition was met.

Conclusions:

• Algebraic Multigrid (AMG) offers a significant computational speedup for the finite element (FE) method in solving the EEG forward problem.
• The findings suggest that AMG-CG can enhance the feasibility of accurate and fast high-resolution FE volume conductor modeling for routine source analysis in EEG.
• This improved efficiency can broaden the application of advanced modeling techniques in neuroscience and clinical diagnostics.