Simulating extravascular artifacts surrounding cortical draining veins to increase neuronal specificity of ultra-high-field fMRI

功能磁共振成像是最常用的非侵入性脑功能成像方法。现代核磁共振成像技术提供了前所未有的高时空分辨率成像方法，使功能磁共振成像研究大大提高了我们对人类大脑工作的认识。所有功能磁共振成像技术都试图通过测量伴随的血液供应变化来推断神经元活动的模式。最常用的fMRI信号是血氧水平依赖或BOLD对比，这通常是通过梯度回声MRI序列获得的。该技术提供了最灵敏和稳健的功能磁共振成像测量，因此非常适合利用可用的时空成像分辨率。然而，限制梯度回声BOLD推断潜在神经元活动能力的一个主要问题是，由于皮质表面引流静脉和皮质内静脉的信号位移所产生的伪影，导致空间特异性的退化。这些伪产物可以利用功能结构的先验知识，结合仔细的实验设计，提供足够的特异性来检测大脑皮质层的神经元激活。“层流fMRI”的前景最近激发了人们对设计方法的兴趣，这些方法通过结合皮层血管化的知识来解释这些限制BOLD fMRI的伪影，这在皮层中是非常规则和一致的，以及fMRI信号如何受到引流静脉影响的模型。在这里，我们建议采用Ogawa/Boxerman模型来模拟血管内和血管外的梯度回波BOLD信号，以广泛适用于去除引流静脉伪影以提高分辨率。我们的新实现将首先根据工件的先前模拟和估计进行基准测试。第二步，对仿真结果进行校正后，建立去除伪影的模型。随后，我们验证了我们的伪影范围和基于体素大小的模型，用于在独立样本中去除伪影。校准和验证将使用在3T和7T时获得的初级视觉皮层中众所周知的地形图的fMRI数据进行。这证明了我们的假设，即考虑血管外信号可以提高fMRI的神经元特异性。目前，一些研究人员正在努力构建场强高达20T的下一代MRI扫描仪，为未来减少fMRI体素大小铺平道路。然而，伪影的范围和幅度随着MR场强的增加而增加。因此，对神经元活动和fMRI测量中观察到的血流动力学信号之间的关系进行仔细的建模对于开发这些未来技术将更加重要。因此，我们将使用我们的模拟来预测空间分辨率和最大可实现的特异性从3到20特斯拉。

fMRI is the most commonly used imaging method for non-invasively mapping brain function in humans. Modern MRI technology has provided imaging methods with unprecedentedly high spatio-temporal resolution, enabling fMRI investigations which are dramatically enhancing our knowledge about the working human brain. All fMRI techniques seek to infer patterns of neuronal activity by measuring the accompanying changes in blood supply. The most commonly used fMRI signal is the blood oxygen level dependent or BOLD contrast, which is typically acquired with gradient echo MRI sequences. This technique provides the most sensitive and robust fMRI measurement and is therefore well suited to exploit the available spatio-temporal imaging resolution. However, a major problem that limits the ability of gradient-echo BOLD to infer underlying neuronal activity is the degradation of the spatial specificity due to artifacts given by signal displacement from draining veins at the cortical surface and from intra cortical veins. Those artifacts can be discounted using a priori knowledge of the functional architecture in combination with careful experimental design, which provided sufficient specificity to detect neuronal activation across the cerebral cortical layers. The prospect of “laminar fMRI” has spurred recent interest in methods designed to account for these artifacts that limit BOLD fMRI by combining knowledge about the cortical vascularization, which is strikingly regular and consistent across the cortex, with models of how the fMRI signal is influenced by draining veins. Here we propose to adapt the Ogawa/Boxerman model to simulate the intravascular and extravascular gradient-echo BOLD signals for a broadly applicable removal of draining vein artifacts to improve resolution. Our new implementation will first be benchmarked against previous simulations and estimations of the artifact. In a second step the model for artifact removal will be constructed after the calibration of the simulation. Subsequently, we validate our artifact range and voxel size based model for artifact removal in an independent sample. Calibration and validation will be carried out using fMRI data of a well-known topographic map in the primary visual cortex acquired at 3T and 7T. This serves as a proof for our hypothesis that accounting for the extravascular signal can improve the neuronal specificity of fMRI. Currently, several endeavors are underway to construct next generation MRI scanners with field strengths up to 20T, paving the way for future decreases in fMRI voxel sizes. However, the artifact ranges and magnitudes increase together with MR field strength. Thus, careful modeling of the relationship between neuronal activity and the observed hemodynamic signals underlying fMRI measures will be even more important to exploit these future technologies. Therefore, we will use our simulations to predict the spatial resolution and the maximum achievable specificity from 3 to 20 Tesla.