Non-invasive imaging of neural activity with magnetic detection electrical impedance tomography (MDEIT): a modelling study.
Non-invasive imaging of neural activity with magnetic detection electrical impedance tomography (MDEIT): a modelling study.
复制标题
使用磁检测电阻抗断层扫描 (MDEIT) 对神经活动进行非侵入性成像：一项建模研究。
DOI：
10.1088/1361-6579/ad0358
复制
发表时间：
2023
影响因子：
3.2
通讯作者：
Mason K
中科院分区：
文献类型：
--
作者：
Mason K
目的(1)开发三维快速神经磁阻抗断层成像(MDEIT)的计算流水线；(2)确定恒流或恒压对MDEIT更有利；(3)在具有真实噪声的人头模型中进行模拟神经活动的重建，并将MDEIT与EIT进行比较；(4)利用光泵磁强计(OPM)在盐水池中进行MDEIT的二维研究并比较重建算法。使用重建扰动的位置和大小的误差来比较图像。二维MDEIT是在一个带有阻性扰动和一个OPM的盐水池中进行的。利用重建扰动的位置和大小误差对6种重建算法进行了比较。主要结果在COMSOL多物理中开发了计算流水线，将雅可比计算时间从几个月减少到几天。对于电流噪声、电流源噪声、电流源噪声和磁强计噪声三种噪声情况，MDEIT重建图像的重建误差比EIT低，平均误差分别为7.0%、5.5%和11%。秩次分析的结论是，MDEIT雅可比比EIT雅可比更少秩亏。在盐水池中重建模型的最佳重建误差为13%，使用0阶Tikhonov正则化和模拟基于噪声的校正来实现。意义本研究证明了用于神经成像的三维MDEIT是可行的，并且MDEIT重建的图像优于EIT，这可以用MDEIT雅可比矩阵的较小秩差来解释。对盐水池中的摄动进行了重建，证明了用OPMS进行二维MDEIT的原理，并确定了最佳重建算法。
Objectives(1) Develop a computational pipeline for three-dimensional fast neural magnetic detection electrical impedance tomography (MDEIT),(2) determine whether constant current or constant voltage is preferable for MDEIT,(3) perform reconstructions of simulated neural activity in a human head model with realistic noise and compare MDEIT to EIT and (4) perform a two-dimensional study in a saline tank for MDEIT with optically pumped magnetometers (OPMs) and compare reconstruction algorithms.ApproachForward modelling and image reconstruction were performed with a realistic model of a human head in three dimensions and at three noise levels for four perturbations representing neural activity. Images were compared using the error in the position and size of the reconstructed perturbations. Two-dimensional MDEIT was performed in a saline tank with a resistive perturbation and one OPM. Six reconstruction algorithms were compared using the error in the position and size of the reconstructed perturbations.Main resultsA computational pipeline was developed in COMSOL Multiphysics, reducing the Jacobian calculation time from months to days. MDEIT reconstructed images with a lower reconstruction error than EIT with a mean difference of 7.0%, 5.5% and 11% for three noise cases representing current noise, reduced current source noise and reduced current source and magnetometer noise. A rank analysis concluded that the MDEIT Jacobian was less rank-deficient than the EIT Jacobian. Reconstructions of a phantom in a saline tank had a best reconstruction error of 13%, achieved using 0th-order Tikhonov regularisation with simulated noise-based correction.SignificanceThis study demonstrated that three-dimensional MDEIT for neural imaging is feasible and that MDEIT reconstructed superior images to EIT, which can be explained by the lesser rank deficiency of the MDEIT Jacobian. Reconstructions of a perturbation in a saline tank demonstrated a proof of principle for two-dimensional MDEIT with OPMs and identified the best reconstruction algorithm.