Computer login: labuser2 BR@ins08 PREPROCESSING Removing - PowerPoint PPT Presentation

Computer login: labuser2 BR@ins08

PREPROCESSING Removing ArFfacts: CondiFoning Data: • spike detecFon • rescaling / inten. norm. • slice Fming • temporal filtering • moFon • spaFal smoothing func,onal ,meseries correcFon • B0 unwarping preprocessed func,onal MODEL FITTING General linear model using: design - Canonical HRF matrix - FIR ‘sFck’ funcFons - Other basis funcFons parameter es,mates Single‐subject analysis & sta,s,cal maps Plots STATISTICAL INFERENCE contrast Thresholding: FWE, FDR, GRF vector ‘ac,va,on’ maps

‘raw’ k‐space data PREPROCESSING Removing ArFfacts: CondiFoning Data: RECONSTRUCTION • spike detecFon • rescaling / inten. norm. • slice Fming • temporal filtering • moFon • spaFal smoothing func,onal ,meseries correcFon • B0 unwarping REGISTRATION / NORMALIZATION preprocessed func,onal ROI high res. anatomical MODEL FITTING General linear model using: design standard template - Canonical HRF extracted matrix - FIR ‘sFck’ funcFons ,mecourse - Other basis funcFons parameter es,mates Single‐subject analysis & sta,s,cal maps Plots STATISTICAL INFERENCE contrast Thresholding: FWE, FDR, GRF vector ‘ac,va,on’ maps

Slice‐Timing WHY WHEN Different slices are collected at debatable: some say not • • different Fmes in TR necessary, as effects are small – may be more helpful if averaging across slices BEST: with moFon correcFon • otherwise: some say before • moFon correcFon, some say a\er HOW • FSL ‘sliceFmer’ or GUI, SPM GUI

MoFon CorrecFon WHY WHEN If a subject moves during (or typically always, unless using a bite‐ • • between) scans, then a chunk of bar Fssue may move to a different voxel before filtering, smoothing, a\er • spike correcFon HOW • notoriously difficult • FSL: mcflirt or GUI, SPM: realignment, mrTools

SpaFal Filtering (smoothing)b WHY WHEN increase SNR by averaging noise if expected acFvaFons are • • of nearby voxels relaFvely large increase likelihood of overlap in a when averaging across subjects in • • standard space a volumetric space Gaussian random fields and other NO: ROI or individual subject • • correcFons require ‘smooth’ data (debatable) Before temporal filtering • HOW • convoluFon w/ a Gaussian kernel • FSL fslmaths or GUI, SPM smoothing

Temporal Filtering / Detrending WHY WHEN Increase model fit by removing low‐ almost always • • frequency changes not related to as a last step • protocol except perhaps for ‘resFng state’ • or other scans which look at slowly varying signals, or if doing a frequency analysis HOW • high‐pass filtering: gaussian vs. Detrending convoluFon in Fme • matlab scripts? • FSL fslmaths –bpd, GUI • inclusion in model • SPM: in model specificaFon • MODEL is also filtered

Model Fifng • General Linear Model (GLM): weights data Y = Xβ + ε error ε ~ N n (0, Vσ 2 ) explanatory variables = * + Worsley, Keith J. (2001) Ch. 14: StaFsFcal analysis of acFvaFon images. In Func%onal MRI: An Introduc%on to Methods. Jezzard, Mathews, Smith ed. Oxford.

= * + O X sFmulaFon hemodynamic response funcFon (HRF)

ROI Analysis • ExtracFng data from an ROI: – increases signal‐to‐noise (assuming the voxels are actually doing the same or similar thing) – generally avoids the mulFple comparisons problem – also avoids the need for normalizaFon across subjects – results in a massive data reducFon • write simple scripts! plot it!