系統識別號 U0026-2507201109191100
論文名稱(中文) 應用統計方法降低像臉實驗的功能性核磁共振資料之噪音干擾
論文名稱(英文) A Study of Noise Extraction on the fMRI Analysis of Two Face Related Experiments
校院名稱 成功大學
系所名稱(中) 統計學系碩博士班
系所名稱(英) Department of Statistics
學年度 99
學期 2
出版年 100
研究生(中文) 陳威仁
研究生(英文) Wei-Jen Chen
學號 r26981041
學位類別 碩士
語文別 英文
論文頁數 148頁
口試委員 口試委員-龔俊嘉
中文關鍵字 噪音  Finite Impulse Response Median Hybrid Filter  獨立主成份分析  Kullback-Liebler Divergence  迴歸分析  功能性核磁共振影像 
英文關鍵字 Noise  Finite Impulse Response Median Hybrid Filter  Independent Component Analysis  Kullback-Liebler Divergence  Regression Analysis  Functional Magnetic Resonance Imaging 
中文摘要 根據 Lindquist (2008) 所提出來的文獻中得知, 一般線性模型(GLM:Generalized Linear Model)是在分析大腦功能性核磁共振(fMRI:functional Magnetic Resonance Imaging)資料簡單且常用的方法。在大腦功能性核磁共振實驗中, 儀器會量測受試者的血氧濃度相依對比(BOLD:Blood Oxygenation Level Dependent)反應。這些反應除了夾帶實驗者所感興趣的訊號(Signal)外, 其中包含最大部分的是一些不知名因素的噪音(Noise), 這些高比例的噪音會造成在建立一般線性模型(GLM)時, 真正的訊號看不見。因此, 為了能夠降低血氧濃度相依對比反應(BOLD response)中噪音的含量, 我們提供了兩個降低噪音的方法有效地凸顯出真正的訊號。進行完個別分析(Individual Analysis)後,我們參考了 Sui 等人在 2009 年提出的文獻進行群體分析(Group Analysis),文中使用了獨立主成份分析(ICA:Independent Component Analysis)和 Kullback-Leibler divergence (K-L divergence)的方法比較兩個不同的群體大腦活化反應區域的差異。

我們發現噪音的干擾往往會造成量測到的血氧濃度相依對比反應有低頻, 中頻或高頻的走勢。在此論文中, 我們比較了兩個降低噪音的方法。第一個方法是我們所提來的, 利用``鄰近相似趨勢的立體像素(Voxel)" (SPINNV:Similar Pattern In Near Neighbor of the Voxel)尋找噪音干擾的走勢。我們假設噪音的成因具有區塊性區域性質, 所以想藉助鄰近相似趨勢的立體像素找出噪音的干擾走勢。我們所提出來的方法在尋找高頻率的噪音有不錯的效果。第二個方法是利用``簡單Finite impulse response Median Hybrid (FMH) 濾波器"尋找噪音的干擾走勢。結果發現, 此方法在尋找中頻或低頻的噪音走勢時有不錯的效果。以上所提到的兩個方法會被用在兩筆成功大學認知所--龔俊嘉 教授實驗室所提供的大腦功能性核磁共振資料。在論文的最後, 我們還進行了模擬分析以證實我們所提出的方法是有效的。
英文摘要 According to Lindquist (2008), the generalized linear model (GLM) is a simple and commonly used method for fMRI analysis. Because the blood oxygenation level dependent (BOLD) responses are often contaminated by unknown interference, the BOLD signal of interest usually will be masked by unidentified noise. Therefore the GLM analysis may have the trouble to find the activated regions. In this study, we considered two methods to reduce unknown interferences in order to enhance BOLD signal. Then, following the approach by Sui et al. (2009), we collect the activated areas through the independent component analysis (ICA) and the Kullback-Liebler divergence (K-L divergence). The methods help us to find out the different activated areas between male and female groups.

The main idea of this study is to notice that the noises may be with low, medium or high frequency. In order to reduce these noises, we proposed two methods for catching the unknown interference. The methods are ``similar pattern in near neighbor of the voxel" (SPINNV) and the simple ``finite impulse response median hybrid (FMH) filter". We assume that certain patterns of noise are similar for the voxels close to each other, so that the idea of SPINNV is to catch the noises by the tracks of local voxels. It turns out to catch the high frequency noise. On the other hand the MHF filter is more effective on removing the medium and low frequency noises. The methods are applied to the data of two experiments about stimuli of face from the Institute of Cognitive Science, National Cheng Kung University (Prof. Kung's lab). A simulation illustrates that our proposed methods are effective.
論文目次 1. Background and Motivation ...... 1
1.1 Data Description and Experimental Procedure ...... 5
1.1.1 Introduction of fMRI ...... 5
1.1.2 Case One ...... 9
1.1.3 Case Two ...... 18
1.2 Research Problems and Proposed Methods ...... 28
1.2.1 Research Problems ...... 28
1.2.2 Proposed Methods ...... 28
2. Literature Review ...... 30
2.1 The Statistical Analysis of fMRI Data ...... 30
2.2 Group-Discriminative Techniques ...... 31
2.3 Software and Package ...... 32
3. Methods ...... 33
3.1 Process to Analysis fMRI Data ...... 33
3.1.1 Generalized Linear Model ...... 33
3.1.2 Locally Weighted Scatterplot Smoothing ...... 34
3.1.3 Independent Component Analysis ...... 35
3.1.4 Kullback-Liebler Divergence ...... 37
3.2 New Methods to De-noise ...... 38
3.2.1 Usage of Similar Pattern in Near Neighbor ......39
3.2.2 Usage of Noise Filter ...... 40
4. Applications to Real Data Sets ...... 42
4.1 Case One ...... 42
4.1.1 Results after De-noise by Using Near Neighbor ...... 42
4.1.2 Results after De-noise by Using Filter ...... 71
4.2 Case Two ...... 83
4.2.1 Results after De-noise by Using Near Neighbor ...... 83
4.2.2 Results after De-noise by Using Filter ...... 107
4.3 Comparison of Methods in Two Cases ...... 119
4.3.1 The Different Characters of the Two Data Sets ...... 119
4.3.2 The Difference in Using near Neighbor ...... 122
4.3.3 The Difference in Using Filter ...... 122
4.4 Simulation ...... 123
4.4.1 Preparing ...... 123
4.4.2 Results One ...... 131
4.4.3 Results Two ...... 138
5. Conclusions and Feature Work ...... 145
5.1 Conclusions ...... 145
5.2 Feature Work ...... 146
Bibliography ...... 147
參考文獻 [1] Buracas, G. T., and Boynton, G. M. (2002), “Efficient Design of Event-Related fMRI Experiments Using M-Sequences”, NeuroImage, 16, 801–813.
[2] Churches, O., Baron-Cohen, S., and Ring, H. (2009), “Seeing face-like objects: an event-related potential study”, NeuroImage, 20, 1290.
[3] Davies, P.L., Fried, R., and Gather, U. (2004), “Robust signal extraction for on-line monitoring data”, J. Statistical Planning and Inference, 122, 65–78.
[4] Dragomir S. C. (2008), “Some Properties for the Exponential of the Kullback-Leibler Divergence”, Tamsui Oxford Journal of Mathematical Sciences, 24, 141–151.
[5] Fried, R., Bernholt, T., Gather, U. (2006), “Repeated Median and Hybrid Filters”, Computational Statistics and Data Analysis, 50, 2313–2338.
[6] Hyvarinen, A., and Oja, E. (2000), “Independent Component Analysis: Algorithms and Applications”, Neural Networks, 13, 411–430.
[7] Hadjikhani, N., Kveraga, K., Naik, P., and Ahlfors, S. P. (2009), “Early (N170) activation of face-specific cortex by face-like objects”, Neuroreport, 20, 403.
[8] Heinonen, P., and Neuvo, Y. (1988), “FIR-median hybrid filters with predictive FIR substructures”, IEEE Transactions on Acoustics, Speech, and Signal Processing, 36, 892–899.
[9] Jiang, F., Dricot, L., Blanz, V., Goebel, R., and Rossion, B. (2009), “Neural Correlates of Shape and Surface Reflectance Information in Individual Faces”, Neuroscience, 163, 1078–1091.
[10] Kalli, S., Heinonen, P., Turjanmaa, V., and Saranummi, N. (1985), “Identifying patterns and profiles in 24-hour blood pressure recordings”, In: Proceedings of the Fifth International Symposium on Ambulatory Monitoring, 113–118.
[11] Kao, M. H., Mandal, A., Lazar, N., and Stufken, J. (2009), “Multi-objective optimal experimental designs for event-related fMRI studies”, NeuroImage, 44, 849–856.
[12] Lindquist, M. A.(2008), “The Statistical Analysis of fMRI Data”, Statistical Science, 23, 439–464.
[13] Loffler, G., Yourganov, G., Wilkinson, F., and Wilson, H. R. (2005), “fMRI Evidence for the Neural Representation of Faces”, Nature Neuroscience, 10, 1386–1390.
[14] Malach, R., Reppas, J. B., Benson, R. R., Kwong, K. K., Jiang, H., and Kennedy,W. A., et al.(1995), “Object-related activity revealed by functional magnetic resonance imaging in human occipital cortex”, Proceedings of the National Academy of Sciences of the United States of America, 92, 8135.
[15] Siegel, A. F. (1982), “Robust regression using repeated medians”, Biometrika, 68, 242–244.
[16] Sui, J., Adali, T., Pearlson, G. D., and Calhoun V. D. (2009), “An ICA-based method for the identification of optimal FMRI features and components using combined group-discriminative techniques”, NeuroImage, 46, 73–86.
[17] Valentine, T. (2001), “Face-space models of face recognition”, Computational, geometric, and process perspectives on facial cognition: Contexts and challenges. Mahwah: LEA, 83–113.

  • 同意授權校內瀏覽/列印電子全文服務,於2016-08-12起公開。
  • 同意授權校外瀏覽/列印電子全文服務,於2016-08-12起公開。

  • 如您有疑問,請聯絡圖書館