光束構成法之腦部活動相關性時空造影

Abstract

大部分的研究都同意腦部的神經元會集體有著同步的活動特徵。過去的研究指出時序上的相關性與腦部區域間的溝通息息相關。在腦磁圖儀與腦電圖儀的研究中，一般會透過神經活動的振盪訊號來研究區域間的功能關聯性。然而，神經活動的時序訊號一般都是包含跨頻帶的資訊，所以從一般性的同步來研究區域間的功能相關性是很重要的。

在這篇論文中，我們提出了一個以光束構成法為基礎的方法來估算腦部活動相關性。透過我們的方法，可以揭露神經互聯網的活動情形。神經互聯網間是利用類似的時序特徵訊號來交換資訊。一但使用者指定了一個區域，我們便可分析該區域中最具代表性的特徵訊號，並找出具有相似特性的區域的分布情況。原則上，將腦中的位置兩兩用我們的方法計算，就可以找出所有可能的神經互聯網。

我們的方法利用最大相關性準則來最大化參考區域和全腦的其他區域的活動相關性。透過這個準則，我們可以用閉形式解的方式正確地解析出訊號源電偶極方向，進而有效的決定對應各個位置的空間濾波器。我們的方法可以計算出相關性分布圖，從這個圖可以看出與參考區域有顯著相關性的腦部區域。

實驗證明了我們的方法確實可以正確的計算出腦部中有相關活動的區域。與過去傳統的腦部定位方法不同的是，我們專注在找出有著與參考訊號類似的時序訊號特徵的腦部皮質區域。除此之外，我們也將方法應用到實際由人腦量測的實驗證明我們方法的可行性。在鏡像神經元實驗中，大多我們找出有相關性的區域都曾在之前情緒處理、臉部感知和鏡像神經原系統的研究中發現。除此之外，我們還可以提供這些與神經互聯網相關的區域在時間上的資訊。

簡言之，我們提出的方法是利用腦磁波來研究腦部活動間的相關性。並建立一個腦部活動動態的時空造影與顯示系統。使用者給定一個神經網路中的區域來當作參考的區域，我們的方法可以推估在每個時間點與參考的區域相關的區域分布情形，進而揭露出這個時間網路的活動情況。

It has been widely accepted that neurons in the human brain collectively have synchronous patterns of activities. The past findings have suggested that temporal correlation may relate to the communications between the distributed areas. There are some studies in magnetoencephalography and electroencephalography that analyze the functional connectivity between cortical areas using the oscillatory features of neuronal activity. However, temporal dynamics of neuronal activities is generally consisted of cross-frequency components. Therefore, it is also important to directly investigate the functional connectivity as well as general synchronization.

In this thesis, we have proposed a beamformer-based imaging method of correlated brain activities that can reveal the neural network with similar temporal patterns for information exchange. The method can identify the correlation distribution referred to a specified position, called the reference region. In principle, we can apply our method on all pairs of grid points to identify all possible neural networks of correlated activities.

Our method exploits a maximum-correlation criterion that maximizes the significant level of correlation between the reference region and the entire brain volume. The maximum correlation criterion helps to analytically and accurately determine the dipole orientation in a closed-form manner and thus determine the spatial filter very efficiently for each position. The correlation map can be calculated to reveal cortical regions with significant similarity to the reference position in the brain.

The experiments with simulation data demonstrated that our method can accurately determine the correlated regions. Different from the conventional source localization method, we focus on the areas which have the similar temporal patterns with the reference signal. We demonstrated the applicability of the proposed method on real data. In the mirror neuron experiment, most of the regions we revealed are reported by the previous findings of emotional processing, face perception and the mirror neuron system. Moreover, we can provide the time information about when these regions are correlated to the neural network.

In summary, the proposed method can be used to directly study dynamics of correlation brain areas based on electromagnetic recordings of brain activities. Given the reference region as one of the areas in the neural network, our method can estimate the correlated regions at each time point and thus reveal the dynamic behavior of the neural network.