Department of Mathematical Analysis, Ghent University, Galglaan 2, B-9000 Ghent, Belgium.
Int J Numer Method Biomed Eng. 2013 Mar;29(3):363-79. doi: 10.1002/cnm.2510. Epub 2012 Sep 29.
It is well known that the uncertain knowledge of the conductivity values of the head tissues has an important impact upon the accuracy of the electroencephalogram source reconstruction. Assuming a certain value of the conductivity often leads to high reconstruction error values when solving the inverse problem. It is possible to quantify the impact of multiple uncertain conductivity values on the localization accuracy. We propose an approach that reduces the impact of these multiple uncertainties on the reconstruction accuracy of the dipole parameters. This paper elaborates the numerical method and shows results of localization accuracy in a five-shell spherical head model. Sensitivity analysis, when considering multiple layers in the head model, shows the different scales of the influence of the various uncertain conductivity values on the potential values. We propose a cost function that reduces the impact of multiple uncertainties of the conductivity value on the electroencephalogram dipole reconstruction and two strategies for selecting potential values on the basis of the sensitivity analysis. Numerical simulations, when considering multiple uncertainties in the model, provide results with higher reconstruction accuracy compared with the case where only a single uncertainty is taken into account.
众所周知，头部组织电导率值的不确定性对脑电图源重建的准确性有重要影响。在解决逆问题时，假设电导率的某个值往往会导致高的重建误差值。可以量化多个不确定电导率值对定位精度的影响。我们提出了一种方法，可以降低这些多个不确定性对偶极子参数重建精度的影响。本文阐述了数值方法，并展示了在五层球形头部模型中的定位精度结果。在考虑头部模型中的多层时的灵敏度分析表明，不同不确定电导率值对电位值的影响具有不同的尺度。我们提出了一种成本函数，可以降低电导率值的多个不确定性对脑电图偶极子重建的影响，并提出了两种基于灵敏度分析选择电位值的策略。在考虑模型中的多个不确定性时的数值模拟，与仅考虑单个不确定性的情况相比，提供了更高的重建精度结果。
