FMRI Time Series Analysis with the Software SPM99

In recent years the technique of functional magnetic resonance imaging (fMRI) has been rapidly developed and improved. Its interest is moving from the mere technical side to wide clinical applications. Functional magnetic resonance imaging yields images with very high resolution and contrast. But its certainly greatest advantage is, to locate human brain activity in a complete non invasive manner, i.e., without the use of external contrast agents. It uses blood, which serves as intrinsic contrast agent, due to its oxygenation state and flow. Image signal changes are induced by the different magnetic susceptibility of oxygenated and deoxygenated blood hemoglobin and the changing blood flow. Those signal changes are in the range of a few percent and cannot be seen with the naked eye. To obtain reliable information about size and location of the activated brain regions, statistical analysis is necessary. This thesis describes the analysis of fMRI images with the software SPM99, where SPM stands for Statistical Parametric Mapping. Firstly, a number of optional preprocessing steps of the software are explained, which can improve the image quality. Then the time course of a single voxel of the images is modeled. The general linear model and theoretical results will be summarized since the statistical models used by SPM99 are all special cases of the general linear model. Temporal filtering is introduced and critically assessed, because it is a possibility to deal with the temporal autocorrelations of the error terms in fMRI time series. In order to be able to report size and location of activated regions, statistical inference is required. Various hypotheses that can be specified and tested using contrasts of the parameter estimates, are listed. Those tests are performed at each voxel, hence a multiple comparison correction is essential. SPM99's multiple comparison correction procedure, which refers to results from the theory of Gaussian random fields and takes into account the dependence of neighboring voxels, is outlined. Finally, various simple activation data sets, where periods of activation and rest alter, are described and analyzed with SPM99.

[1]  R Baumgartner,et al.  Quantification of intensity variations in functional MR images using rotated principal components. , 1996, Physics in medicine and biology.

[2]  C. Crawford,et al.  Optimized gradient waveforms for spiral scanning , 1995, Magnetic resonance in medicine.

[3]  I. Wilkinson,et al.  Introduction to functional magnetic resonance imaging , 1999 .

[4]  Karl J. Friston,et al.  To Smooth or Not to Smooth? Bias and Efficiency in fMRI Time-Series Analysis , 2000, NeuroImage.

[5]  Karl J. Friston,et al.  Analysis of fMRI Time-Series Revisited—Again , 1995, NeuroImage.

[6]  Andrew P. Holmes,et al.  Statistical issues in functional brain mapping. , 1994 .

[7]  Stuart Clare,et al.  Functional magnetic resonance imaging : methods and applications , 1997 .

[8]  Karl J. Friston,et al.  Analysis of functional MRI time‐series , 1994, Human Brain Mapping.

[9]  Karl J. Friston,et al.  Characterising brain images with the general linear model , 1997 .

[11]  Karl J. Friston,et al.  Statistical parametric maps in functional imaging: A general linear approach , 1994 .

[12]  Karl J. Friston Analysing brain images: principles and overview , 1997 .

[13]  Karl J. Friston,et al.  Analysis of fMRI Time-Series Revisited , 1995, NeuroImage.

[14]  Karl J. Friston,et al.  Spatial transformation of images , 1997 .

[15]  R. Turner,et al.  Functional magnetic resonance imaging of the human brain: data acquisition and analysis , 1998, Experimental Brain Research.