Review Flashcards
What are the effects of the magnetic fields in MRI?
- Static magnetic field –> atoms align with the direction of the magnetic field
- Oscillating magnetic field = radio frequency pulse
–> Spins of individual atoms get in phase
–> Atoms flip and take the direction of the oscillating field –> increase in energy state (angle of flip)
If have static mag field and apply RF pulse, knock over protons but want to address parts of the brain so the RF field interacts with protons in terms of their resonance, we can select slices by adding gradients. We dont have just a static field, we add gradients. Only the protons where there is resonance will be affected by the RF pulse
RF pulse knocks the p+ off and when RF stops the p+ come back and when they do they emit É which is capted by the antennas
Phase encoding gradient: speeds up the spin of p+ so when it stops the p+ are at a diff phase so we can look at that pt in space
Frequency encoding gradient
MRI: take 3D space and apply 3 gradients to select pixels
What is a gradient magnetic field in MRI?
- As linear as possible, stationary and of short duration
- Slice selection gradient: applied at time of RF pulse
- Phase encoding gradient: use of de-phasing after RF pulse
- Frequency encoding gradient: applied at time of read out signal
How do the physical principles in MRI give rise to an image with anatomical structure?
- Emitted signal decays over time
- Signal intensity depends on several factors:
-> Density of H+
-> T1-recovery: recovery of longitudinal orientation = spin-lattice relaxation
-> T2-decay: loss of transverse magnetisation due to the loss in phase coherence = spin-spin interactions
Releases radio f when goes back to relaxation state, capted by antennas - Factors are different in different tissues, resulting in signal contrast
What is a CT scan?
Some of the gamma rays are absorbed more by some
tissues than others
Take lots of X-rays from many diff angles and computer
software puts them all together
MRI gives better contrast but is much more expensive
What is magnetic resonance spectroscopy?
Can look at specific markers in the brain, shine radiation and get spectrogram of what bounces back, depending on the f of what bounces back get idea of what the structure is
Areas of resonance that correspond to diff mol types
What is DTI?
- Pattern of action potentials
-> From where and to where
-> Paths = axons
See degree to which the flow of water is anisotropic (not random)
More myelination = more anisotropic - Large bundles of 1000s of axons
- Axons that start and end in each other’s vicinity, stay together –> white matter pathways or tracts
Can look at regions of interest and see to where and to what extent they are connected
What is fMRI?
- BOLD signal: Hb contains iron so is ferromagnetic,can see how much oxy blood is in a certain area of the brain at a certain pt in t, can see how much É a task needed and how much an area of the brain is active depending on the task
- Deoxy Hb: magnetic momentum (paramagnetic)
-> Alters spin-spin interactions –> faster T2 decay
-> Increase in oxygenation –> increased fMRI signal - More macroscopic side effects of paramagnetic particles: field inhomogeniety and tissue susceptibility
–> Total dephasing = T2* decay
BOLD is a relative response, we always have to compare it to ex a baseline
What are the 3 components of HRF?
- Initial dip: decrease in blood oxygenation and measured signal
The O2 is used by tissues then the vasc rapidly rép to the need in O2 - Primary (strongest) response: influx oxygenated blood –> strong increase in signal (in response to the metabolic demand)
- Negative overshoot: signal decreases
Have t series of what we,re recording in the brain that shows us this fction
BOLD fction has a canonical shape so we can leverage this info to extract this fction from our recordings, fit the response to the t series, line up all the slices taken over t with the shape of the hemodynamic rép
What is the problem with multiple comparisons?
- Whole brain analysis around 30,000 tests –> high expected rate of false positives –> need to correct for multiple comparisons
- Can have a ROI a priori or do corrections
- Bonferroni correction: desired probability of type I/# of comparisons (# of voxels)
-> Too conservative: correction needed for # of independent tests
-> Nearby voxels: joined signal and noise during data acquisition, further correlations introduced during pre-processing (spatial smoothing), thus nearby voxels don’t provide independent tests
What is correlational MVPA?
Test whether within-condition correlation between datasets is higher than between-condition correlation
What is decoding MVPA?
- Across-voxels activity patterns in dataset 1 used to train a pattern classifier
- Classifier finds a decision boundary in the multi-dimensional input space
- Cross-validation: test classifier performance on independent dataset 2
What is fNIRS?
- Apply infrared light which will interact with the Hb, the amount of light that bounces back is proportional to the amount of Hb in the area that we shined light on
- Sensitive to outer cortical region of the brain
- Based on similar principles to MRI
- Measure change in Hb, from which neural activity is inferred
- Light source emits light at 2 specific wavelengths
- Photons bounce off in all directions and some photons make it back to the detector
- The ratio between the 2 f that come back tells you the amount of oxy and deoxy Hb there is
- The 2 wavelengths are absorbed differently depending on the concentration of oxy and deoxy Hb
- Can estimate the concentration of oxy and deoxy Hb between source and detector
- Channels in fNIRS are made of optodes which are between a light source and a light detector
Why does dendritic architecture matter?
For the signal to be abe to be captures at the surface of the scalp, the sum of the PSPs has to be big enough
Measure voltage change across the scalp or the mag fiels associated with is
Mostly record pyramidal cells in the 5th layer of the cortex bc the have the right architecture
EEG signals need to be amplified and digitised
What is an ERP?
The brain responds to events, record signal continuously, play a stim and then use the timing info to epoch the data (chunk it into segments around the time of the stim) and average them together so we can remove the noise that we’re not interested in, the noise is random and the signal is related to the stim, the phases o the noise will cancel out bc they’re random
Averaging increases our signal to noise ratio
What are the properties of the field signal?
The field activity oscillates in time so the signal is represented as a time wave. The wave can also be represented as a rotation. Properties of oscillation like frequency, phase and amplitude are used to describe and analyse the signals
