This article introduces a unique and powerful new method for spatial localization of neuronal sources that exploit the high temporal resolution of magnetoencephalography (meg) data to locate the originating sources wi...
详细信息
This article introduces a unique and powerful new method for spatial localization of neuronal sources that exploit the high temporal resolution of magnetoencephalography (meg) data to locate the originating sources within the brain. A traditional frequency beamforming algorithm was adapted from its conventional application to yield information on the spatial location of simulated neuronal signals. The concept is similar to that used in signal sourcelocalization in magnetic resonance imaging (MRI) in which spatial location is determined by the frequency of oscillation of the MR signal. Whereas a traditional frequency beamformer uses the time course values of all sensors in the dataset to assign a power value for each possible frequency in the signal, it provides no information on the spatial location of those frequencies. Our approach assigns a power value to each location in the three-dimensional head volume. To compute this power value, the time courses of a subset of sensors closest to that location in space are used rather than all the time courses in the dataset. Our novel technique incorporates actual meg sensor locations of the closest sensors at each location in space. The approach is relatively simple to implement, yields good spatial resolution, and accurately spatially locates a simulated source in low signal-to-noise environments. In this work, its performance is compared to that of the synthetic aperture magnetometry (SAM) beamformer and shown to exhibit improved spatial resolution. (C) 2015 Elsevier Ltd. All rights reserved.
A matching pursuit (MP) based algorithm, called source deflated matching pursuit (SDMP), is proposed for locating sources of brain activity. By iteratively deflating the contribution of identified sources to multiple ...
详细信息
ISBN:
(纸本)9781424441228
A matching pursuit (MP) based algorithm, called source deflated matching pursuit (SDMP), is proposed for locating sources of brain activity. By iteratively deflating the contribution of identified sources to multiple measurement vectors (MMVs), the SDMP algorithm transforms the original multi-basis-vector/matrix selection problem into a single-basis-vector/matrix selection problem, which not only mitigates the residual-source interference but also remedies the intrinsic bias when locating deep sources. The robustness of the proposed algorithm to two bias factors is verified through simulations.
Recent dynamic sourcelocalization algorithms for the Magnetoencephalographic inverse problem use cortical spatio-temporal dynamics to enhance the quality of the estimation. However, these methods suffer from high com...
详细信息
ISBN:
(纸本)9781457717871
Recent dynamic sourcelocalization algorithms for the Magnetoencephalographic inverse problem use cortical spatio-temporal dynamics to enhance the quality of the estimation. However, these methods suffer from high computational complexity due to the large number of sources that must be estimated. In this work, we introduce a fast iterative greedy algorithm incorporating the class of subspace pursuit algorithms for sparse sourcelocalization. The algorithm employs a reduced order state-space model resulting in significant computational savings. Simulation studies on meg source localization reveal substantial gains provided by the proposed method over the widely used minimumnorm estimate, in terms of localization accuracy, with a negligible increase in computational complexity.
This thesis introduces the usage of non-convex basedregularizers to solve the underdetermined meg inverse *** to be reconstructed is considered to have a structure whichentails group-wise sparsity and within group spa...
详细信息
This thesis introduces the usage of non-convex basedregularizers to solve the underdetermined meg inverse *** to be reconstructed is considered to have a structure whichentails group-wise sparsity and within group sparsity among itscovariates. We discuss the usage of ?2 normregularization and smoothed ?0 (SL0) normregularization to impose group-wise and within group sparsityrespectively. In addition, we introduce a novel criterion which ifsatisfied, guarantees global optimality while solving thisnon-convex optimization problem. We use proximal gradient descentas the method of optimization as it promises faster convergencerates. Initially, we show that our algorithm successfully recoverssparse signals with a smaller number of measurements than theconventional ?1 regularization framework. Wealso support this claim using meg source localization simulationsand extend the reconstruction for both stationary andnon-stationary signals. Next, we formulate a global convergenceanalysis for the novel algorithm. Finally, we incorporate novelinformation criteria techniques and concepts of duality to find thebest set of regularization parameters and a proper stoppingcriterion respectively. We were able to successfully illustratethat the regularization parameters (models) with lower informationcriteria performs better than the ones with higher informationcriteria. Also, concepts of duality provides the necessary tools todetermine when to stop the algorithm, which is an importantcontribution considering the non- differentiability of theobjective function
In this article, the relationship between motion artifacts and magnetoencephalography (meg) sourcelocalization errors was established for the meg system based on array optically pumped magnetometers (OPMs). The influ...
详细信息
In this article, the relationship between motion artifacts and magnetoencephalography (meg) sourcelocalization errors was established for the meg system based on array optically pumped magnetometers (OPMs). The influences of bias fields on OPM's performances were first evaluated through theoretical and experimental analysis, whose results proved that the bias magnetic fields along the sensitive axis affect the gain and sensitivity of OPM, while the pumping axis bias magnetic fields affect the direction of the sensitive axis. The OPM with a large linewidth has better anti-interference ability but worse sensitivity. Then, meg simulations were carried out on the basis of OPM performance testing. The relative error of the forward model proved that the rotated sensitive axis would affect the forward model, while the gain error and sensitivity directly affect the signal-to-noise ratio (SNR) of the meg signal. The experiment and simulation also proved that it is recommended to use biaxial mode for OPM-meg experiments and the motion artifact detected by OPMs should be reduced to less than 1nT at least. Moreover, the linewidth of the OPM's cell is positively correlated with the anti-interference capability of OPM. Therefore, sensitivity and anti-interference capability must be simultaneously considered in the design of OPM by optimizing the cell's linewidth to adapt to more scenarios.
暂无评论