fmri Flashcards
PET: how does it work
you inject a tracer into bloodstream, this takes 30s to go to brain where it is active, rCBF, it decays and emits a positron which collides with an electron and photons are emmitted showing as gamma rays, takes 30s to peak.
PET methodological issues
- requires blocked design
- can only measure once due to radiation
- but can study near cavities (unlike fmri)
- temp freq: minutes
BOLD signal is the actual signal
HRF is the model fitted to the signal
BOLD size is arbitrary
- dip: more Hb -> more distortion by paramagnetic Hb -> less signal. ~2s
- overcompensation: less Hb, more HbO2 -> less distortion by the paramagnetic Hb, more signal. ~4 to 8s
- undershoot: more Hb. ~8 to 10s
How does mri scan work
- alignment of protons to MF by the big magnet
- pulse with Larmor frequency to produce resonance of the precession
- 90 degree knockdown by RF pulses (head coil), producing net vector perpendicular to MF that is detected by head coil
- relaxation a) T1 weighted uses longitudinal relaxation
b) T2 and T2* weighted uses spin spin transversal relaxation, phase decoherence. Hb paramagnetic is quicker to have decoherence.
- specific pulses wiht varying TR (time until next excitation) and TE (time of measurement)
blocked design
- no randomized trials
- cant infer specific processes timing
- often not compatible with cognitive tasks
- can better detect small differences, more power
- doesnt take as much time
- easy to analyze
- good for state based processes
- may be better if event related difficult for patient grp
event related design: 12-15 s per trial
- randomized trials, can vary ISI
- less sensitive head motion
- takes longer
- switching between trial types can sometimes lead to confusion
- better for attention of the pp
- can correlate a specific process during the task
- must average over a timelocked event that must be assigned post-experiment/depends on pps judgement
- cant use if event is infrequent
fMRI methodological issues
- cant distinguish excitation/inhibition
- activity observed may be sufficient to lead to bhvr, but not necessary for the bhvr
- susceptibility artifacts: voids
- acoustic artifacts: scanner noise: must wait before next scan bc BOLD could be response of auditory features
- temporal drifts
- head motion
- smoothing
- brain normalization with atlas
- multiple testing correction
fmri spatial and temporal res
tradeoff: bigger boxels is worse spatial res but chance of bigger activation, whereas smaller voxels better spatial res and less partial volume effects (sometimes a voxel is on a border between different tissue types which reduces contrast)
spatial: < 1 mm3 till 6mm3. limited by: microvascolature, MF, sequence, coils, instrumentation, voxels
temporal: hundreds of ms/ till 1 or 2 s: limited by HRF characteristics
mri designs
- subtraction (needs pure insertion): (A+B) - (B) = A [and AB]
a) conjunction: min 2 conditions with same component, look at common activation of 2 contrasts. (A+B )- (A) = B and AB. (B+C)-(C) = B and B*C. Common activation gives you approximation of B.
b) factorial: focus on differneces. AB/CD. (A+B) - (C+D) = main effect factor row, (A+C)-(B+D) gives main factor column. (A-B)-(C-D) gives interaction. - parametric
- information based mapping
disagreeing imaging vs lesion data?
i) lesion yes, fmri no: - experimental task and baseline both use the process
- activity in region hard to detect by fmri
- lesion may damage tracts from another region
ii) lesion no, fmri yes: dependent on strategy used?
- the activation is a general process
- the region is inhibited
- not enough power in lesion study
rapid event related: 2 - 6s per trial
- jittered ISIs (4s)
- overlapping BOLD, but can deconstruct IF linear additivity
- better due to not long ISIs (so that random thoughts dont come in between)
- can do more trials so also better power
mixed design
- rest blocks
- active blacks with events
Trojano experiment: visuospatial imagery vs perception
- imagery: which time has their clock hands further apart, time given by audio
- perception: while viewing clocks onscreen
- control: is syllable number in both times odd or even?
blocked design
results: in imagery vs syllabes contrast PPC was activated but not in imagery vs perception so there is some overlapping process involved
Wagner study of recall memory: event vs blocked design
investigate what area is correlated with succesful remembering of words in a recognition test after list learning.
1. blocked: can only use conditions that we know enhance memory: semantic (abstract, concrete word?) vs nonsemantic processing.
semantic - nonsemantic contrast: LIFG, parahippocampal, fusiform
2. event: look at activity during encoding of those trials that were correctyl remembered (high confidence hits and misses only) : same results
explanation: time-on-task? no
how does localization work
- gradient coils vary strength MF locally in xyz directions
- complex wave is picked up, fourier analysis on it can then tell the relative strength of the precession speeds to get the strength of the signal in that location
