Figure 10.

To examine the estimation accuracy of numerical LFP solutions under differing current source handling methodologies (shifting vs. splitting) a range of models with varying resolution were generated, solved, and quantitatively compared with an analytical estimate via the line-source equation in the plane at 50 μm under the slice model, and then with each other. Red regions represent voxels where shifting results in more error relative to splitting. μ, σ in each window denote the average and standard deviation of residual differences over all voxels and all time-steps. While splitting often outperformed shifting, particularly in tetrahedral meshes at very low resolutions, as meshes became more complex, shifting, and splitting methods reached comparable levels of error (low single digit μV). The difference in methods across the range of resolutions was more pronounced in hexahedral meshes for any given node density because of the relative element size between hexahedral and tetrahedral meshes.