Department of Molecular and Integrative Physiology, University of Kansas Medical Center, Kansas City, KS 66103, USA.
IEEE Trans Biomed Eng. 2010 Jul;57(7):1652-62. doi: 10.1109/TBME.2010.2047858. Epub 2010 Apr 19.
Nonparametric iterative algorithms have been previously proposed to achieve high-resolution, sparse solutions to the bioelectromagnetic inverse problem applicable to multichannel magnetoencephalography and EEG recordings. Using a mmse estimation framework, we propose a new algorithm of this type denoted as source affine image reconstruction (SAFFIRE) aiming to reduce the vulnerability to initialization bias, augment robustness to noise, and decrease sensitivity to the choice of regularization. The proposed approach operates in a normalized lead-field space and employs an initial estimate based on matched filtering to combat the potential biasing effect of previously proposed initialization methods. SAFFIRE minimizes difficulties associated with the selection of the most appropriate regularization parameter by using two separate loading terms: a fixed noise-dependent term that can be directly estimated from the data and arises naturally from the mmse formulation, and an adaptive term (adjusted according to the update of the source estimate) that accounts for uncertainties of the forward model in real-experimental applications. We also show that a noncoherent integration scheme can be used within the SAFFIRE algorithm structure to further enhance the reconstruction accuracy and improve robustness to noise.
先前已经提出了非参数迭代算法，以实现适用于多通道脑磁图和 EEG 记录的生物电磁逆问题的高分辨率、稀疏解。我们使用最小均方误差估计框架，提出了一种新的此类算法，称为源仿射图像重建（SAFFIRE），旨在降低对初始化偏差的敏感性，增强对噪声的鲁棒性，并降低对正则化选择的敏感性。所提出的方法在归一化导联空间中运行，并使用基于匹配滤波的初始估计来对抗先前提出的初始化方法可能产生的偏差影响。SAFFIRE 通过使用两个单独的加载项来最小化选择最合适的正则化参数的困难：一个固定的、依赖于噪声的项，可以直接从数据中估计，并且自然地源自最小均方误差公式；一个自适应项（根据源估计的更新进行调整），用于在实际实验应用中考虑正向模型的不确定性。我们还表明，可以在 SAFFIRE 算法结构中使用非相干积分方案，以进一步提高重建准确性并提高对噪声的鲁棒性。
