Lecture 6: Fundamentals of fMRI - Part 2 Preprocessing Flashcards
What is intersubject variability? - (2)
Variability in fMRI data across a set of subjects; it
includes the factors associated with intrasubject variability, along with between-subjects differences in task
performance and physiology.
e.g., one participants’ area of brain may react quicker than same brian in another participan
What is intrasubject variability?
- Variability in the fMRI data from a single subject
associated with thermal, system, and physiological noise, as well as with variability in the pattern of brain activity during task performance.
What does this diagram show? - intersubject and intrasubject variability in hemodynamic response in participants
- Here, hemodynamic responses derived from four different brain regions are plotted for 20 subjects within the same experiment. - (5)
- Although there are substantial differences between
individuals in the timing of their hemodynamic responses, with some peaking earlier and some later than the canonical hemodynamic response (lines), - there is relatively good consistency in the timing of the hemodynamic response across regions within the same individual. (After Handwerker et al., 2004.)
- show general pattern which is good which is needed for analysis - (standard) canovical HRF is used for analysis
- In some areas of the brain, some participants’ responses does not allign too well with standard (canovial )HRF than others - inter subject
- Variation between people and variation in different parts of brain in participant e.g., subject 4 each part of brain has different hemodynamic response
Diagram - providing evidence that within subject is better than between subject + consistency of responses - (5)
- Dave (on left) going in both as experimenter and participant in MRI scanner for a very long period of time and doing tasks being exposed to visual stimuli
- This study wanted to see whether Dave would have the same BOLD signal responses on different occasions
- In A, show patterns of BOLD signal varied over time with voxel signifiance threshold being strict
- In B, when lowered voxel significance threshold, see similar BOLD signal activation across each occasions shown in each task
- The concept is that within a given pp and given brain region, we see consistent responses in hemodynamic response in given task –> fMRI data provides reliable link to underlying neural activity
Due to inter-subject and intra subject variation, within-subject design is needed as… - (2)
- If we did between-subject and looked at different people’s brain we will be comparing very unlike things potentially
- Big advantage statistically of using within subject of seeing how the same brain area of the same participant behaves in different tasksif possible
What is the key characteristics of BOLD signal? - (4)
- Sensitive
- Reliable
- (Approximately) Linear
- Sluggish
The characteristic of BOLD signal being sensitive is that
Visual stimulus as brief as 5ms can produce measureable signal (Yeşilyurt et al., 2008)
The characteristic of BOLD signal being reliable means - (3)
- Responses found across the whole brain (grey matter)
- Consistent within-subject and region
- BUT Varies between subjects and brain regions
The characteristic of BOLD signal being approximately linear means - (2)
- Signals scale with stimulus intensity/neural activity
- Signal from successive stimuli sum together (where stimulus durations > several seconds) - at least if keep stimulus/task changes far enough apart in time
The characteristic of BOLD signal being sluggish means
Response takes around 5s after stimulus onset to peak and up to 15 s to return to baseline.
The implications of characteristic of BOLD signal for experimental design - (5)
- Inter-subject variation means that between subject designs are less favourable and typically avoided unless essential to the research question.
- Within subjects designs are favoured where possible – they make use of the reliability of BOLD signals measured repeatedly in the same part of the same person’s brain.
- The sluggish nature of hemodynamic responses relative to neural activity means that fMRI experiments have a natural tempo.
- Task manipulations occur at a rate designed to elicit large hemodynamic changes over several seconds.
- The (approximate) linearity of the BOLD signal on this timescale means that signals elicited by different tasks can to some extent be “unpicked” and decomposed into different neural activity using linear models (GLM, next week).
To increase ampltidue of BOLD signal is to increase the
scanner field strength (T - Telsa)
As static field strength increase in scanner, the - (2)
raw BOLD signal changes r increases quadratically
Thus 3.0T scanners measure 4x as much of raw signal than 1.5 T
What has Turner and colleagues found of effects of field strength in fMRI - (2)
- Turner and colleagues measured changes in visual cortex activation at 1.5 T and 4.0 T.
- Approximately 2 to
3 times as much signal was recorded at the higher field strength..
Greater field strength has 2 main effects on spatial distribution of activation - (3)
- Improvements in spatial specficity e.g., 7.0T used and identify spatial patterns of activity that reflect ocular dominance columns
- Result in better sensitivty to signal changes from blood vessels
- Increases spatial extent of activation - numbers of voxels within a region
What happens to T1 and T2 when increasing magnetic field strength (T) in fMRI? - (2)
- The parameter T1 increases with field strength (by about 30% from 1.5 T to 3.0 T), and this could reduce the effective signal recovery for short TR values.
- The parameter T2* decreases with increasing field strength,
and this could reduce the time available to acquire a signal.
What is suspectibility artefacts?
Signal losses on T2* dependent images due to magnetic field inhomogenities in regions where air and tissue is adjacent (i.e., around sinuses - nose)
The problem with high-field MRI is that - (4)
- Suspectibility artifacts disorts the uniformity of magnetic field
- Suspectibiltiy artiefactts increase at higher field strengths at air tissue boundaries
- Both images scanned at 1.5 T (A) and 4.0 T (B)
- B shows areas of singal loss in ventral lobe in darker patch and more extensive in 4.0 T than 1.5 T
To minimise the unwanted variability in BOLD signals nearly all fMRI studies incorporate
preprocessing algortihms which compensate for specific sources of variability (e.g., head motion)
What is preprocessing? - (2)
computational processing procedures that are applied to fMRI data that follow image reconstruction but before statistical analyses
Preprocessing steps are intended to reduce variability in the data that is not associated with experimental task to prepare the data for statistical testing
Example of signal to noise - (2)
Signal = freind’s speech
Noise = other sounds that interfere with your ability to hear friend
What is signal in fMRI? - (2)
Meaningful changes in some quantity
For fMRI an important class of signals include changes in intensity associated with BOLD resposne across a series of T2* images
What is noise in fMRI? - (2)
Nonmeaningful changes in some quantity.
There are many sources of noise in fMRI studies, and some changes may be classified as either noise or signal, depending on the goals of the study.
We can improve ability to detect fMRI signal by - (2)
increasining its ampltidue or decreasing its noise
Thus increasing signal-to-noise ratio
What is signal to noise ratio?
relative strength of signal compared with other sources of variability in the data
The quantity measured in fMRI which is MR signal reflects both changes in
- net magnetisation caused by excitation pulse (signal) and fluctuations caused by thermal energy in sample and imaging hardware (noise)
When is raw SNR used?
evaluate the performance of the scanner hardware, and institutions compare such measures of SNR when
deciding which MRI scanner to purchase and when monitoring the quality of an MRI scanner over time.
Contrast of MRI image refers to
physical property to which it is sensitive
For example, an image senstivie to T1 contrast will be… - (2)
- An image sensitive to T 1 contrast will be bright for voxels with short T1 values (like white matter) and dark for voxels with long T1 values (like gray matter or cerebrospinal fluid).
- Thus,T 1 weighted pulse sequences have a good ability to distinguish between gray matter and white matter (high CNR), but only a limited ability to distinguish between cerebrospinal fluid and air (low CNR).
CNR stands for contrast-to-noise ratio which is
magnitude of the intensity differences between quantities divied by variability in their measurements
Functional MRI relates the
changes in brain physiology over time to an experimental manipulation
Structural images of fMRI provide
snapshot of underlying tissue
MRI data is collected as
ime series so examining changes associated with experimental tasks like voxels increases/decreasing
What is a session? - (2)
a single visit to scanner by subjectt
For fMRI studies, each session usually includes both structural and functional scans
What is a run?
An uninterpreted presentaiton of experimental task usually lasting 5 to 10 minutes
Typical T2* weighted images have low structurl contrast making it hard to
identifty boundaries between different types of tissues
The important quantity for fMRI studies is functional signal to noise ratio which is - (2)
The ratio between the intensity of a signal associated with changes in brain function and the vari- ability in the data due to all sources of noise.
Functional SNR is sometimes
called dynamic CNR or functional CNR.
What are the five main causes of temporal and spatial nosie in fMRI? - (5)
- Intrinsic thermal noise within the subject and the scanner electronics
- System noise associated with imperfections in the scanner hardware
- non-neural biological sources: Artifacts resulting from head motion, respiration, heart rate, and other physiological processes
- Variability in neuronal activity associated with non-task-related brain
processes - Changes in behavioral performance and cognitive strategy.
Instrinsic thermal noise, system noise (scanner hardware) , non-neural biological soruces (e.g., head motion, respiration…) can be mitgated by
preprocessing
Changes in neural activity due to non-task related processes and uncontrolled changes in behaviour and cognitive stratgery can by mitigated by
task and experimental design
All MR imagning whether anatomical or funcitonal is subject to thermal noise/intrinsic noise which is
fluctuations in MR signal intesnity over space or time that is caused by thermal motion of electrons within the sample or scanner hardware
The thermal nosie increases linearly with
field strength that is approximately twice as large in 3.0 T as to 1.5 T
Magnitude of thermal noise within a voxel is indepndent of its
spatial location
Thermal noise is dependent on voxel’s
signal ampltiude
What is system noise?
fluctuations in MR signal intensity over space or time that is caused by imperfect functioning scanner hardware
One source of system noise is scanner drift which is
slow changes in voxel over time
The common cause of scanner drift is
changes in resonant frequency of hydrogen protons associated with subtle changes in strength of magnetic field
What is physiological noise? - (2)
Fluctuations in MR signal intensity over space and time due to physiological activity of human body
Sources of physiological noise include motion, respiration, cardiac activity and metabloic reactions
Signal variability due to subject motion is common and extremely disruptive for fMRI studies.
Throughout an experiment, a subject
may shift the position of her head , move shoulders, arms, legs or become more comfortable or swallow ebcause of nervousness
Motion due to cardiac activity in most fMRI studies is
too fast to be sampled effectively (i.e., TRs greater than 500ms)
Respiration introduces variability in fMRI signal through systematic disortions in magnetic field as
- As the subject breathes, the expansion of the
lungs casts a magnetic susceptibility “shadow,” influencing field strength and homogeneity of the magnetic field and altering signal intensity throughout the image (including areas outside the brain).
Diagram of noise in fMRI - (4)
- (A) noise from all sources
- (B) physiological noises due to variation in blood flow and merabolic processes
- (C) noise due to bulk head motion and cardiac and respiration (D)
- (E) and (F) show data collected from fluid-filled ball and can’t escape from noise (F)
Physiological nosie raather than thermal/system noise is dominant source of
variability in fMRI studies especially at higher field strengths
There is also changes in activity due to non-task related processes being noise such as
planning events, recalling memories, thinking of appoitments later
There is behavioural and cognitive variability in task performance as becoming noise in fMRI studies - (5)
- There is Intersubject and intrasubject variability in reaction/response time
- Variability in response time has consequences in timing and ampltiude of BOLD signal
- Performance, aside from RT and response time, can be measured by exaimining accuracy of responses.
- However, accuracy is frequently related to response time in participants can perform most tasks more accurately when doing it slowly = known as speed-accuracy trade off
- There is stragery changes another source of task-induced variability, in cognitive task there might be one stratgery to use but others adopt other stratgeries resulting in individual differences in activation - solve it by doing pilot studies
What does this diagram show - reproducibility of fMRI activations across sessions in one participant - (4)
- Participant attended 33 fMRI sessions each containingmotor, visual and cognitive tasks
- Experimental procedures were repeated in same manner
- In A, it shows different patterns of activations evoked in different sessions based on voxels passing convential sig criteria of p < 0.05
- In B, analysis showed there was activation differences found between sessions was large
Most labs to prevent head motion is
head restraints (A), (B) bite bar, (C) thermoplastic molds to participants’ head or not shown in image is participate in training sessions with MRI scanner in mock scanner
When the head moves during an experiment, some of the images will be
acquired with the brain in the
wrong location
What is the goal of motion correction?
Adjust the series of images so the brain is always in the same position
Preprocessing and mitigating noise can be solved by - (4)
Slice time correction
Motion correction
Spatial smoothing
Temporal filtering
Slice time correction - (4)
TRs (time needed to acquire an entire volume) for fMRI are typically non-negligible (e.g., 2-3s)
This means data are acquired in one slice of the brain (e.g., the top) much earlier relative to task induced HRF than in another slice (e.g., the bottom)
Temporal interpolation is one method used to correct this (looking at nearby timepoints to estimate the amplitude of the MR signal at a single point in time for the whole volume)
Temporal interpolation is used to work out for each slice in a given volume what would have been the activity at a single point in time –> middle of volume what is HRF is?
What is the definition of temporal interpolation?
The estimation of the value of a signal at a time point
that was not originally collected, using
data from nearby time points.
What does this diagram show? - effects of head motion on fMRI data - (4)
(A) Following a discrete movement of the head, large-intensity transitions exist at tissue boundaries,
including the edges of the brain.
(B,C) The effects of head motion on voxel
intensity. The magnified views show the position of the brain before head motion (B) and after a movement of one voxel to the right (C).
The numeri- cal intensity values for the voxels within the blue square are shown below.
Note that the intensity in a given voxel may change by more than a factor of 5 due solely to head motion.
What does this diagram show? edge effects of head motion in fMRI analysis - (5)
Edge effects of head motion in fMRI analyses. Because the intensity transitions in the brain are greatest at its edges, head motion often results in systematic rings of artifactual activation around the edges of the brain.
Shown are activation maps, in four axial slices, derived from the analysis of a motor task
Here, however, the brain moved forward by two voxels at the beginning of one stimulus onset and remained for-
ward for 4 s.
The result was large changes in signal intensity around the edges in the brain; for example, voxels at the back of the brain changed from high intensity to low intensity as a result of the movement.
Thus, these voxels had a significant decrease in activation (plotted in a blue-to-white color map) due entirely to the
movement, and these movement-related effects dwarf the significant task-related increases in activation (plotted in a red-to-yellow color map) within motor cortex.
What does these diagram show - plots of head mtion over an experimental session - (7)
Plots of head motion over an experimental session.
Shown are plots of translational (A) and rotational (B) head motion from a single fMRI session.
By convention, translational effects comprise movements from left to right (x-axis), forward to back (y-axis), and top to bottom (i.e., in the slice acquisition direction; z- axis).
Rotational effects comprise turns around the x-axis (pitch), around the y-axis (roll), and around the z-axis (yaw).
This experiment consisted of seven runs, each of
410 images, with a TR of 1500 ms.
Large motions between runs are visible as verti-
cal lines on the plot; for example, see the vertical line near image 2050 that reflects an upward through-plane movement.
Note that these are estimated motion values
at each point in time and thus can be influenced by a number of factors besides
head motion itself, as indicated in the text.
Motion correction - (3)
With a very rich sequence of images, it is not difficult to estimate the amount of movement in the time series.
These movements are corrected by neuroimaging software so that all the images line up as closely as possible.
This can still leave some problems: 1) the data have to be resampled to fix the motion and 2)the data in a plane of the resliced images were no longer acquired at the same time
Spatial smoothing - (3)
Many sources of noise have a small spatial scale, whereas BOLD signal changes tend to operate on a larger scale.
Spatial smoothing works by “averaging” the signal in each voxel with its neighbours.
Noise tends to cancel out, while signal is boosted.
What does this diagram show? - temporal filtering
Noise from the pulse and breathing produces much higher frequency signal changes than signal changes induced by task changes (here 20s blocks of motor activity alternating with 20s rest)
What does this diagram show?
Other sources of noise can be very low frequency, e.g., “scanner drift” in which MR signal gradually waxes and wanes over many minutes.
Diagram of raw signal to high pass filtered
Temporal filtering - (2)
Temporal filters are used to remove MR signal fluctuations associated with scanner drift (low frequencies removed by a high-pass filter) and physiological processes (high-frequencies removed by a low-pass filter, or by acquiring data a low frequency)
By filtering noise above and below the frequencies associated with task manipulations the Signal to Noise Ratio is increased
Which of the following determines the shortest timeescale that can be resolved in fMRI?
A. participants capacity to remain immobile and attentive
B. Timing of tasks and stimuli chosen by the experimenter
C. duraiton and linearity of hemodynamic responses to neural activity
D. spatial specificiy of blood supply
C.
To exploit the reliability of BOLD signals measured repeatedly in the same part of the same person’s brain, fMRI experiments should, where possible:
A.use event-related design
B.use between-subjects design
C.use within-subjects design
D. organize timing of stimuli/task so processes of interest are minimally correlated over time
C
Which preprocessing step is most useful in mitigating the effects of “scanner drift”:
A.applying a low-pass temporal filter
B.motion correction
C.applying spatial smoothing
D. applying a high-pass temporal filter
D.
Which of the following describes “superposition”, characteristic of a linear system?
A.Magnitude of the response scales with the magnitude of the input
B.Responses to more than one input correspond to the sum of the responses to the individual inputs
C.BOLD signal increases as the concentration of oxygenated hemoglobin increases
D.T2* signal increases as the concentration of oxygenated hemoglobin increases
B
