fMRI and BOLD Responses

fMRI: Applied Research Methods

• Functional MRI (fMRI) is a technique used to map brain activity by measuring the amount of oxygen in the blood.
• fMRI uses the Blood Oxygenation Level-Dependent (BOLD) signal.
• fMRI is sensitive to changes in blood flow in response to neural activity.
• The BOLD response is slow (around 1-4 seconds).

What is fMRI?

• fMRI measures brain activity by detecting changes in blood flow associated with neuronal activity.
• fMRI uses the principle of magnetic resonance imaging (MRI) to acquire images.
• fMRI measures the subtle changes in the magnetic properties of hemoglobin, specifically deoxyhemoglobin.
• fMRI is based on the principle that blood flow increases to areas of the brain that are active, and this increased blood flow contains more oxygenated hemoglobin, which affects the magnetic properties of the brain.

The fMRI Experiment

• The fMRI experiment begins with a research question.
• Experimental designs vary (e.g., block design or event-related design).
• Block design involves grouping experimental conditions into alternating blocks in time.
• Event-related design presents stimuli as individual events.

Interpreting fMRI Data

• Data analysis focuses on identifying patterns in the BOLD response.
• Data analysis involves identifying correlations, relationships, and differences between conditions.
• Data analysis helps to interpret brain activity changes in response to stimuli, conditions or tasks.

MRI vs fMRI

• MRI studies brain anatomy.
• fMRI studies brain function.

fMRI Quiz

• fMRI has a particularly good temporal resolution (1-3 seconds) and spatial resolution (1-3mm).
• EEG is not related to fMRI.

MRI Examination Steps

• The participant is placed in a magnetic field.
• A radio wave pulse is sent at a specific frequency.
• The participant emits a signal, which is measured.
• The emitted signal is used to create an image of the tissue.

Magnetic Resonance Imaging (MRI) Scanner Components

• Static magnetic field
• Radiofrequency coil
• Gradient coils

Physics of MRI

• The scanner is a large electromagnetic coil.
• It is made superconducting by cooling with liquid helium (-261 °C).
• It produces a very strong static magnetic field (e.g., 3 Tesla).
• Protons within the brain can be affected by the main magnetic field of the scanner.
• The spinning of protons around their axes creates a magnetic field.
• Protons align with the magnetic field during the process called magnetization.
• Hydrogen protons align with the magnetic field.
• Radiofrequency pulses change the alignment of protons, which generates signals for image acquisition.

T1 and T2 Relaxation Times

• T1: time needed for the aligned hydrogen protons to return to their original position after stimulation - This varies in different brain tissues.
• T2: time taken for the protons to lose their radiofrequency energy and regain equilibrium - This also varies in brain tissues.

Measuring Human Brains

• T1 relaxation measures the return to the aligned state of hydrogen protons.
• This differs in gray matter, white matter and cerebrospinal fluid.
• T1 measures the time taken to return to equilibrium along the longitudinal magnetization axis.

Functional Imaging (measuring brain activity)

• The brain uses 20% of the body's total oxygen consumption.
• An increase in neural activity results in an increased blood flow and oxygen to neurons to meet metabolic demands.
• fMRI is responsive to the amount of oxygen in the blood.
• A network of blood vessels carry glucose and oxygen throughout the body, including the brain.
• Hemoglobin brings glucose and oxygen to the neurons.
• Deoxyhemoglobin (hemoglobin without oxygen) alters local magnetic fields.
• Distortions in the magnetic field are proportional to the amount of deoxyhemoglobin, showing differences in brain activity.

BOLD Signal (Blood Oxygenation Level-Dependent)

• BOLD response is the indirect measure of brain activity
• BOLD signal shows the change in magnetic field, due to oxygenation levels.
• The BOLD signal shows how blood flow adjusts to the increase in oxygen metabolism that occurs during neural activity.
• A positive BOLD response involves an initial dip, an overshoot and a post-stimulus undershoot.

Hemodynamic Response Function (HRF)

• Change in BOLD over time. Responsiveness is rather slow (1-4 seconds).
• BOLD is recorded repeatedly in the brain slices.

Anatomical MRI (T1) vs. Functional MRI (T2)

• MRI (T1) has high resolution (1 mm).
• fMRI (T2) has a low resolution (2-3 mm).
• MRI (T1) produces a 3D volume.
• fMRI (T2) produces a series of 3D volumes.
• MRI (T1) image weights the liquid component as dark and solids as white to determine the structures.
• fMRI (T2) image weights the active regions as white, and passive regions as black, to show how the regions are active.

fMRI in Research Papers

• Key terminology for understanding fMRI research papers (subjects, sessions, runs, volume, slices, voxels).

Experimental Design: Presentation Patterns

• Block designs present conditions in alternating blocks over time.
• Event-related designs present individual stimulus events and analyze neural activity in response to each.

Subtraction method

• Subtraction method derives time needed for cognitive processes from reaction times.
• The method is utilized in behavioral experiments in the cognitive sciences.
• A classical example is how long it takes the brain—first distinguishing between colors (e.g., red/green) and second selecting a motor response (e.g., left/right hand).
• Key assumption in subtraction method is that processes are independent of each other (in other words they do not interact).

Factorial Design

• This design considers multiple factors at once to overcome limitations of the pure insertion assumption.
• A way to illustrate this is by asking about why you can't tickle yourself.
• If self-produced touch sensations are reduced compared to externally produced touch sensations, it is because the brain can predict the tactile sensation from the motor command.
• This is often referred to as 'efference copy.'

Parametric Design

• Parametric design assesses associations between brain activity and function, focusing on the relationship between varying levels, not just differences between conditions.
• Examples include assessing how brain activity relates to the speeds of speech, by testing out passive listening to speech with varying speeds.

Repetition Suppression

• Repeated stimulation reduces neural response.
• This technique presents parts of stimuli twice and measures the decrease in response to the second occurrence of the stimulus.

Inter-Subject Correlation

• This assesses how similar brain responses are across individuals.
• Similarity in reactions is higher for more meaningful stimuli, in higher-order regions.

Within-subject Encoding Models

• Encoding models use natural speech stimuli, identifying brain regions that engage with different aspects of the speech (such as visual, or auditory or contextual details), allowing for understanding of brain activity based on the content of the speech.

Data Pre-processing

• Raw fMRI data is noisy.
• Preprocessing techniques aim to minimize noise from non-task-related variability in the raw data, including correcting for head motion, heartbeat, and breathing during the task.
• Removing unwanted variations in the magnetic field.
• Correcting for individual brain differences.

fMRI Analysis Steps: Modeling the BOLD Response

• To find which voxels respond to the tasks or stimuli.
• A two-step approach: First level analysis (within-subject) and Second level analysis.
  • First level: using techniques like General Linear Model (GLM) in each voxel analyzing data for each participant.
  • Second level: combining data of the participants to compare the response during tasks or conditions.

Predicting the BOLD Response

• Using the shape of the BOLD response to individual events (HRF) and the time points of events in the experiment (experimental conditions).
• These are combined to derive a prediction of the BOLD response.

General Linear Model (GLM)

• The GLM constructs a model to predict the measured BOLD signal by combining experimental conditions as regressors (variables).
• The predicted BOLD signal is a linear combination of these regressors.

General Linear Model: Parameter Estimation

• Beta weights are the optimal parameters for the linear combination in the GLM that best explain the data.
• The GLM finds the optimal values, to find the best fit to predict BOLD data with high accuracy.

General Linear Model: Statistical Inference

• Once the GLM is done, the beta-maps can be generated, that include the beta values for each voxel during the experimental conditions.
• Comparing beta maps of different conditions are combined using statistical tests (e.g. t-tests or ANOVAs) to determine differences between conditions (activation from house > face etc).

fMRI Results: Region of Interest (ROI) Analysis

• ROI-based analysis focuses on specific regions of interest (predefined areas) instead of analyzing the entire brain.
• Whole brain analysis explores differences across the entire brain, while ROI-based analysis zeroes in on particular regions beforehand.

Quick Recap

• Careful pre-processing is important for fMRI analysis.
• The first level analysis involves GLM to find parameter values (beta weights) to describe observed data for each participant.
• Summarise results in beta maps.
• Second level analysis tests data from participants to systematically check conditions
  • statistical tests to determine if there are significant differences in brain activity between the conditions.

Multiple Comparisons Problem

• fMRI experiments often produce many significant results.
• Using simple criteria can lead to false positives.
• Correction methods (like Bonferroni correction) are needed to avoid this problem.

Dead Salmon Study

• The study utilized fMRI technology but did not analyze the significance in the findings (the analysis itself was flawed).

Further Reading

• Mentioned resources with different authors and books to learn and study about Functional Magnetic Resonance Imaging

Assignment

• Data file includes beta values.
• The files contain data from sixteen participants.
• The data involves three regions of interest (ROIs): face-selective, hand-selective, object-selective
• Two ROI sizes were used: 5 and 100 voxels.
• The data includes conditions such as Attend Hand and Attend Face.
• The questions focus on interpreting the findings in the brain areas of interest.

This quiz explores the fundamental concepts of fMRI, focusing on the BOLD signal and its implications in brain activity measurement. Participants will answer questions related to the hemodynamic response, MRI mechanisms, and the physiological properties of human tissues involved in imaging. Test your understanding of the principles behind functional MRI.