L10_SPM_Chap12.ppt.pptVIP

* * * * * * * * * Statistical Parametric Mapping Lecture 10 - Chapter 12 Preparing fMRI data for statistical analysis Textbook: Functional MRI an introduction to methods, Peter Jezzard, Paul Matthews, and Stephen Smith Many thanks to those that share their MRI slides online Figure 12.1. Average fMRI from finger study and its k-space magnitude representation (both for the same slice). k-space values are log10 of raw values Reconstruction of slice-by-slice real and imaginary k-space images was done on the MRI system (Siemens 3T Trio). Raw data can be accessed for k-space processing if needed. k-space processing Ghost correction additional readouts without phase encoding to calculate shifts in readout directions phase/amplitude compensation for physiological noise translation for some registration by applying first order phase shift Slice Timing No within slice correction (100 msec). Slices corrected to time of first slice in the volume. All slices in corrected image volume appear to have been collected simultaneously. Correction often done using phase shift applied following FFT then transforming back to time domain using FFT-1. Provides more accurate interpolation. Assumes slices remain spatially aligned throughout time course motion must be corrected (another chapter) motion induced distortions must be corrected where F(f) is the Fourier transform of f(t), the time series. The phase correction term is straightforward to calculate since the time to shift (dt) is known from pulse sequence information. Spatial Filtering Used to improve SNR signal ranges from 0.5-5% so to achieve SNR of 10:1 need noise to range from 0.05-0.5% of the mean signal. Spatial filtering increases spatial correlation by averaging within a neighborhood of each voxel (kernel extent) Can be implemented in Fourier domain (as multiplication) or spatial domain (as convolution), but spatial domain is simple for small extent filters. Spatial domain filtering is usually separable such that 2D fil