使用鄰位相關效應及自我迴歸分析進行功能性核磁共振影像分析

Abstract

近年來功能性核磁共振影像被當作研究腦功能的重要工具。SPM提供了一個模型導向的分析方式，利用檢定的方式去分析功能性核磁共振影像。ICA提供了另一種資料導向的分析方式。我們將試著用資料導向的方式進行分析，並且解決ICA跟分群方法中高時間複雜度的問題。相關係數在許多功能性核磁共振影像分析中被使用，並且證明是有效的。自我迴歸分析也被應用於多種的時間序列分析上。我們將利用相關係數以及自我迴歸分析進行濾波，將大量被視為雜訊的資料點移除。經過資料的刪減，我們可以進行階層式分群，並且引入相關性演算法。在SPM的資料庫中取得一組聽覺實驗的資料，將我們的分析方法代入此組資料，我們可以正確的擷取到大腦中的聽覺區。

Functional magnetic resonance imaging (fMRI) has become an important tool for brain function studies. Statistical Parametric Mapping (SPM) provided a model-driven method for fMRI studies. Different from SPM, Independent Component Analysis (ICA) provided a data-driven method. Here we are trying to give a data-driven method and solve the high-complexity problem in clustering method and ICA. Correlation coefficient had been used in many ways for fMRI analysis recently and proved efficient. Autoregression analysis is utilized in different time series analysis. Here, we are going to take correlation coefficient and autoregression analysis as two filters, to remove most of the voxels which are considered to be noise. After data reduction, we provide Hierarchical cluster and correlation algorithm to those remaining voxels. Providing the method to the fMRI data from SPM’s auditory experiment, we can correctly get the auditory cortical.