Denoising spinal cord fMRI data: Approaches to acquisition and analysis
Denoising spinal cord fMRI data: Approaches to acquisition and analysis
复制标题
DOI：
10.1016/j.neuroimage.2016.09.065
复制
发表时间：
2017-07-01
期刊：
影响因子：
5.7
通讯作者：
Brooks, Jonathan C. W.
中科院分区：
文献类型：
--
作者：
Eippert, Falk;Kong, Yazhuo;Brooks, Jonathan C. W.
人类脊髓的功能磁共振成像(FMRI)是一项困难的工作，因为脊髓的横截面直径小，由于磁场的不均匀而导致信号丢失和图像失真，以及来自心脏和呼吸源的生理性噪声的混杂影响。然而，由于脊髓作为大脑和身体之间的主要感觉运动界面以及参与各种感觉和运动病理的作用，脊髓功能磁共振成像引起了人们的极大兴趣。在这篇综述中，我们概述了用于解决脊柱功能磁共振成像中的技术挑战的各种方法，重点是减少生理噪声的影响。我们首先描述了为脊柱功能磁共振成像的特殊需求量身定做的采集方法，目标是提高信噪比并减少获得的图像中的失真。接下来，我们将重点介绍解决噪声有害影响的图像处理和分析方法。虽然这些方法包括各种标准的预处理方法，如运动校正和空间滤波，但主要关注的是可应用于基于任务的数据集和静止状态数据集的去噪技术。我们回顾了依赖于外部获得的呼吸和心脏信号的基于模型的方法，以及使用数据本身估计和校正噪声的数据驱动的方法。最后，我们对已成功应用于脑成像降噪的技术进行了展望，这些技术的使用可能有益于人类脊髓的功能磁共振成像。
Functional magnetic resonance imaging (fMRI) of the human spinal cord is a difficult endeavour due to the cord's small cross-sectional diameter, signal drop-out as well as image distortion due to magnetic field inhomogeneity, and the confounding influence of physiological noise from cardiac and respiratory sources. Nevertheless, there is great interest in spinal fMRI due to the spinal cord's role as the principal sensorimotor interface between the brain and the body and its involvement in a variety of sensory and motor pathologies. In this review, we give an overview of the various methods that have been used to address the technical challenges in spinal fMRI, with a focus on reducing the impact of physiological noise. We start out by describing acquisition methods that have been tailored to the special needs of spinal fMRI and aim to increase the signal-to-noise ratio and reduce distortion in obtained images. Following this, we concentrate on image processing and analysis approaches that address the detrimental effects of noise. While these include variations of standard preprocessing methods such as motion correction and spatial filtering, the main focus lies on denoising techniques that can be applied to task-based as well as resting-state data sets. We review both model-based approaches that rely on externally acquired respiratory and cardiac signals as well as data-driven approaches that estimate and correct for noise using the data themselves. We conclude with an outlook on techniques that have been successfully applied for noise reduction in brain imaging and whose use might be beneficial for fMRI of the human spinal cord.