Simulation of Extended Source Localization using sLORETA Method for Magnetocardiography

Abstract

In this study, cardiac source localization was simulated using the spatial filter method. Three types of spatial filters were obtained using the standardized low-resolution brain electromagnetic tomography (sLORETA) method, based on different examination procedures. In Type A filter, the examination was conducted at the front of the torso. In both Type B and Type C filters, the examinations were conducted at the front and back of the torso; however, the distance from the frontal observation plane to the center of the heart model was different for each type. In the simulation experiments, first the goodness of fit (GOF) value was introduced to determine the proper threshold for each spatial filter. Then, single and multiple dipole sources were simulated at different depths with and without noise. The extension of the solutions computed using these spatial filters was investigated. Finally, the performances of these spatial filters, with the corresponding averaged thresholds, were evaluated using the GOF. Type B and Type C spatial filters demonstrated reduction in the extension of source dependency on source depth and improvement in the accuracy of source localization with noisy data.

Imaging the extent and location of spatiotemporally distributed epileptiform sources from MEG measurements

“…In our study, the choice of LCMV threshold was based on both prior literature and empirical testing, aiming to give the optimal geometric mean for comparison. Although more advanced statistical thresholding methods exist, such as Otsu’s threshold ( Chowdhury et al, 2016 ), Bonferroni correction, false discovery rate, and concavity of the survival function ( Maksymenko et al, 2017 ), to the best of our knowledge, the empirical thresholding is still the most common approach for extent estimation in these algorithms ( Palmero-Soler et al, 2007 , de Gooijer-van de Groep et al, 2013 , Attal and Schwartz, 2013 , Sun and Kobayashi, 2017 ). Nevertheless, FAST-IRES still outperformed the benchmark method with a lower spatial dispersion and a higher geometric mean of precision and recall.…”

Imaging the extent and location of spatiotemporally distributed epileptiform sources from MEG measurements

“…In our study, the choice of LCMV threshold was based on both prior literature and empirical testing, aiming to give the optimal geometric mean for comparison. Although other more advanced statistical thresholding methods exist, such as Otsu’s threshold (Chowdhury et al, 2016), Bonferroni correction, false discovery rate and concavity of the survival function (Maksymenko et al, 2017), to the best of our knowledge the empirical thresholding is still the most common approach for extent estimation in these algorithms (Palmero-Soler et al, 2007; de Gooijer-van de Groep et al, 2012; Attal and Schwartz, 2013; Sun and Kobayashi, 2017). Nevertheless, FAST-IRES still outperformed the benchmark method with a lower spatial dispersion and a higher geometric mean of precision and recall.…”