Institut de Neurosciences Cognitives de la Méditerranée, UMR 6193, CNRS & Aix-Marseille Université, Marseille, France.
Neuroimage. 2011 Jan 15;54(2):1196-210. doi: 10.1016/j.neuroimage.2010.08.041. Epub 2010 Aug 26.
Voltage sensitive dye imaging (VSDI) is the only technique that allows to directly measure neuronal activity over a large cortical population. It thus gives access to the dynamics of lateral interactions within or between cortical areas. However, VSDI signal suffers from a weak signal-to-noise ratio and processing methods are either rudimentary or dedicated to spatial or temporal denoising alone. Here we present an innovative method inspired by fMRI data processing, where the goal is to allow, for the first time, denoising of spatio-temporally inseparable VSDI signals and in the most challenging experimental condition, i.e. single trials in awake behaving monkeys. The method is based on a linear model (LM) decomposition of individual VSDI trials. The LM was designed meticulously by identifying all noise and signal components that are known to affect VSDI. We then compared its output against the classical methods based on blank division and detrending. LM proved to be significantly much more efficient to denoise spatial maps and temporal dynamics compared to these usual techniques. It also largely reduced trial-to-trial variability. These performances resulted in a four-fold improvement of signal-to-noise ratio and a two-fold increase of response detectability. Hence, with this method, fewer trials were needed to reach a high signal-to-noise ratio. Lastly, we showed that the LM method can accommodate for a large range of response dynamics, a crucial property for estimating spatial spread of activity or contrast dynamics. We believe that this method will make a strong contribution to imaging dynamics of population responses with high spatial and temporal resolution in trial-based experiments of awake animals.
电压敏感染料成像（VSDI）是唯一一种能够直接测量大皮质群体神经元活动的技术。因此，它可以获得皮质内或皮质间横向相互作用的动态。然而，VSDI 信号的信噪比较弱，处理方法要么是基础的，要么专门用于空间或时间去噪。在这里，我们提出了一种受 fMRI 数据处理启发的创新方法，其目标是首次允许对时空不可分离的 VSDI 信号进行去噪，并且在最具挑战性的实验条件下，即清醒行为猴子的单个试验。该方法基于个体 VSDI 试验的线性模型（LM）分解。LM 通过识别已知影响 VSDI 的所有噪声和信号分量来精心设计。然后，我们将其输出与基于空白分割和去趋势的经典方法进行了比较。与这些常用技术相比，LM 被证明在去噪空间图和时间动态方面要高效得多。它还大大降低了试验间的可变性。这些性能导致信噪比提高了四倍，响应可检测性提高了两倍。因此，使用这种方法，只需较少的试验就可以达到较高的信噪比。最后，我们表明，LM 方法可以适应大范围的响应动态，这是估计活动空间传播或对比动态的关键特性。我们相信，这种方法将为在清醒动物的基于试验的实验中以高空间和时间分辨率成像群体反应动力学做出重要贡献。
