10. From Models to Applications

• human brain very different from other animals

• want to know how communication and cognition works

• with certain techniques we can record data from the whole brain at once

cerebral cortex

profit from ability to communicate and common conceptualisation

• most ns methods are unethical on humans

• indirect measure of neuronal activity

• low spatial and temporal resolution

• large scale measurements

• brain

• systems (inference)

• maps (inference)

• circuits

• neurons (only reduced sets in spec area)

no: synapses and molecules

• EEG

• MEG

• PET

• fMRI

Electroencephalography:

• Measures electrical activity of the brain via electrodes on the scalp; high temporal resolution, low spatial resolution, non-invasive

Magnetoencephalography:

• Detects magnetic fields produced by neural activity; high temporal resolution and better spatial resolution than EEG, non-invasive but expensive

Positron emmission tomography:

• Tracks brain metabolism using radioactive tracers; good spatial resolution, poor temporal resolution, invasive due to radiation

Functional magnetic resonance imaging:

• Measures changes in blood oxygenation (BOLD signal) related to brain activity; high spatial resolution, moderate temporal resolution, non-invasive

• MRI

• CT

• Ultrasound

Uses strong magnetic fields and radio waves to create detailed images of soft tissues; high spatial resolution, non-invasive, no radiation

Uses X-rays to produce cross-sectional images of the body; fast and good for detecting bone, bleeding, or tumors, but involves radiation exposure.

Uses high-frequency sound waves to image soft tissues and organs in real time; safe, portable, and non-invasive, but lower image resolution.

• strong static magnetic field

• Radio frequency or high frequency pulse (amount of energy added depend on how much can be absorbed)

• magnetic gradients

MRI is based on nuclear magnetic resonance (NMR).

In a strong static magnetic field, hydrogen protons align with the field, creating a net magnetization.

The RF pulse adds energy and tips the protons out of alignment. The Larmor frequency determines how much energy is absorbed and depends on the magnetic field strength and the gyromagnetic ratio.

resambles how sensitive the molcule reacts to magnetic field

The RF pulse creates magnetization perpendicular to the static field. When turned off, protons return to equilibrium, emitting a measurable signal (FID).

T1 is the time for longitudinal recovery; T2 is the time for transverse decay. Different tissues have different T1 and T2, influencing image contrast.

Each tissue has specific T1/T2 values, affecting how quickly it emits signal. Signal intensity is modeled with exponential decay/recovery formulas based on T1 and T2.

Gradients vary the magnetic field across space, allowing spatial localization of signals in 3D by encoding position.

A magnetic gradient is applied during RF excitation, exciting only protons in a certain spatial band. The gradient strength determines slice thickness.

During data acquisition, a gradient causes frequency differences across space, letting frequency encode position along one axis.

Phase encoding applies a gradient before signal readout, shifting proton phases to encode spatial info along another axis.

k-space is a matrix storing frequency and phase data from the MR signal. It represents spatial frequencies, not the image itself.

1. Slice Selection – A gradient applied during RF pulse to excite only a specific slice by matching the local Larmor frequency.

2. Frequency Encoding – A gradient applied during signal readout; it makes signal frequency vary by position along one axis (e.g. x-axis).

3. Phase Encoding – A gradient applied briefly before signal readout; it shifts the phase of spins along the perpendicular axis (e.g. y-axis) to encode spatial position.

A 2D Fourier Transform translates k-space data into spatial information, forming the final image from frequency components.

The MRI signal is first stored in k-space as frequency and phase information. A 2D Fourier Transform then converts this data into a spatial image, where signal amplitude determines the brightness of each pixel.

The BOLD signal reflects changes in oxygenated vs. deoxygenated hemoglobin. Active brain regions use more oxygen, altering local magnetic properties and the MRI signal.

The BOLD signal reflects local neuronal processing, especially local field potentials (LFPs), not direct spiking. It shows regional input and processing, related to blood flow changes caused by active neurons.

The HRF describes the typical shape of the BOLD signal after a stimulus: it peaks ~6 seconds later. It is used as a model to analyze fMRI data by convolving it with stimulus timing/ usde as an impulse response.

sensory, motor and cognitive functions take place in individual anatomical regions

GLM identifies brain regions that respond to a stimulus or task by modeling expected BOLD responses (via HRF convolution) and comparing them to the measured signal.

fMRI measures changes in the MRI signal decay (T2*) of hydrogen protons.

Active brain areas receive more oxygen-rich blood, which reduces magnetic disturbances caused by deoxygenated hemoglobin.

This leads to a slower signal decay, resulting in a stronger BOLD signal.

MRI detects this through the faster or slower fading of the proton signal after the RF pulse.

Connectivity between brain regions is what is important for brain function

Resting-state fMRI measures spontaneous brain activity without any task.

It looks for correlated activity patterns between brain regions over time.

Regions that show synchronized fluctuations are considered functionally connected.

The Default Mode Network (DMN) is a set of regions that are active together during rest.