fMRI: Principles and Applications

Lecture 2: What and Why of fMRI

• MRI experiments require specific considerations
• MRI and fMRI operate on distinct physical principles
• fMRI measures brain activity
• fMRI experiments have both advantages and disadvantages
• fMRI is non-invasive and doesn't require opening the skull
• fMRI does not involve radioactivity
• fMRI scans the entire brain relatively quickly and simultaneously
• fMRI utilizes widely available hospital MRI scanners
• fMRI provides functional imaging for cognitive and neuroscience research
• MRI provides structural imaging, is T1-weighted, has high spatial resolution (~1mm), slow acquisition, and reveals brain anatomy
• fMRI provides functional imaging, uses the BOLD signal, has lower spatial resolution (~3mm), high temporal resolution, and reveals brain activity over time
• MRI uses a strong magnetic field and radio waves to align hydrogen protons
• RF pulses excite the protons
• The emitted signal from the protons are measured
• Different tissues generating different signals, ultimately creates an image
• BOLD (Blood Oxygen Level Dependent) signals measure changes in blood oxygenation levels due to said neural activity
• Increased neural activity leads to increased oxygenated blood, leading to an increase of MRI signal
• fMRI's limitations include it being an indirect measure of neural activity with a slow response time of approximately 2-6 seconds
• fMRI is suitable for whole-brain studies, but is sensitive to movement artifacts
• There is a trade-off between spatial and temporal resolution in fMRI

Lecture 3: Subtraction Method and Language Processing

• A brain area "lighting up" in fMRI indicates increased activity
• fMRI requires multiple experimental conditions
• Effective contrasts are vital in fMRI experiments Areas involved in language processing can be identified via fMRI
• fMRI relies on relative activation measures, not absolute ones
• Comparisons between conditions is needed, as the brain is always active
• An example of the subtraction method is comparing brain activation when viewing faces vs. objects to isolate face-selective area
• The BOLD signal is in arbitrary units, meaning there is no universal baseline
• The brain is always active, making it necessary to remove background activity
• Cognitive processes are complex, requiring careful isolation of the process of interest

Language Processing in the Brain

• The classic model involves Broca's Area (Frontal Lobe) for speech production
• The classic model involves Wernicke's Area (Temporal Lobe) for language comprehension
• The modern view involves utilizing the Ventral Pathway for meaning and conceptual encoding in the Temporal Lobe
• The modern view involves utilizing the Dorsal Pathway for speech articulation and grammar in the Parietal and Temporal Lobes
• Studies show a consistent language processing network across different languages, known as the "Universal Language Network"
• Language comprehension and production occur in parallel

Alternative Methods to Subtraction

• Conjunction Analysis identifies regions involved in a shared cognitive process across tasks
• Factorial Design examines interactions between multiple variables
• Parametric Design measures response across varying task difficulty levels
• Adaptation Studies looks at changes in activation with repeated stimuli

Lecture 4: Confounds and Social Conformity

• Challenges are present in fMRI study designs
• Confounds must be defined and recognized in fMRI experiments
• Social conformity must be understood in brain imaging
• A confound is a factor that unintentionally correlates with the experimental conditions, making results difficult to interpret
• Studying maternal attachment using images of a mother's child may be confounded by familiarity as an example

Common fMRI Confounds

• Psychological Confounds include unintended mental processes, like when thinking about lunch during a task
• Behavioral Confounds include differences in task performance influencing activation, like when motor learning is confounded by movement speed
• Methodological Confounds include issues with image acquisition, like visual stimuli confounded with general visual cortex activation

Social Conformity Study in fMRI

• Asch's Conformity Experiment had participants that conformed to incorrect group judgments in perceptual tasks

• An fMRI Study on Conformity had participants perform a mental rotation task with group influence

• Higher activation occurred in the intraparietal sulcus (IPS) when conforming to humans vs. computers

• There was found to be no significant activation in frontal decision-making areas

• Social pressure influences visual processing

• Increased activation might reflect increased cognitive effort rather than altered perception as an explanation for Conformity Activation

• People may also re-check their answers due to social pressure rather than truly perceiving something differently

• Confounds, such as attention to faces vs. computers, can explain the amygdala activation

• Subtraction designs help control for confounds but require careful contrast selection

• Social conformity in fMRI studies must account for alternative explanations beyond perception changes

• fMRI research must carefully design conditions to ensure valid interpretations

Final Summary

• fMRI measures brain function using the BOLD signal, but the BOLD signal limits it's slow temporal resolution and indirect neural activity measurement
• Subtraction designs are useful but require careful control conditions to avoid confounds
• Language processing is distributed across multiple brain areas, with both ventral and dorsal pathways playing crucial roles
• Social conformity studies show brain activation changes, but interpretations must consider alternative explanations like increased cognitive effort

Lecture 5: Statistics & Multiple Comparisons in fMRI

• Understanding voxel time series to activation value transformation
• Ability to identify noise sources in fMRI
• Ability to explore p-values, Type 1 & Type 2 errors
• Intuition for model fitting
• BOLD (Blood Oxygen Level Dependent) signals measure neural activity indirectly
• Data is noisy, and statistical modeling (e.g., General Linear Model - GLM) helps identify meaningful patterns
• Hemodynamic response function (HRF) is convolved with the task predictor

Statistical Considerations in fMRI

• Noise sources include head motion, scanner drift, physiological changes (heart rate, respiration), and measurement errors
• Type 1 error (False Positive) incorrectly detecting activity where none exists
• Type 2 error (False Negative) is failing to detect real activity
• The Multiple Comparisons Problem arises because with thousands of voxels the chance of findings increase
• Correction Methods include Family-wise error rate (FWE), Bonferroni correction, cluster-based correction, and permutation testing
• Beta (β) values represent amplitude of responses in GLM
• Threshold-based statistics involve adjusting significance thresholds to balance false positives and negatives
• A p-value is the probability of observed data occurring under null hypothesis

Lecture 6: Circularity & Multitasking in fMRI

• Understanding region of interest (ROI) analysis
• Recognizing circular analyses and their impact
• Conducting group analyses and correcting for multiple comparisons
• Contrasts between experimental conditions create activation maps
• Group analysis (e.g., paired-sample t-tests) ensures individual variability doesn’t drive findings
• The Multiple Comparisons Problem is analyzed using family-wise error correction, cluster correction, or permutation testing

Circularity in fMRI Analysis

• Circularity is when analysis guarantees a specific result regardless of data
• Selection Bias involves choosing voxels based on the same data used to analyze them
• Sorting Bias involves categorizing voxels before comparison, ensuring a certain outcome
• One example is to select voxels responding to happy faces, and then test their response to other emotions (biased selection)
• A solution is to use independent data for selection and analysis

Multitasking and Learning in the Brain

Hippocampus and Basal Ganglia are two learning systems with different sensitivities to attention and distraction Probabilistic classification tasks are used to study implicit vs. explicit learning systems Circularity in multitasking research is an example of correlating voxels with performance after pre-selecting them, inflating correlations artificially

Lecture 07: Development of Face Processing Areas

• Understanding development of category-selective brain areas
• Investigating innate vs. experience-based neural organization
• Gaining a understanding of evidence from infants, monkeys, and blind individuals
• fMRI studies reveal specialized areas for faces (FFA), places (PPA), bodies, words, etc, which are all examples of Category Selective Areas
• A key question in the field is related to whether or not these areas predetermined (genetics) or experience-driven
• The Retinotopic Protomap Hypothesis (visual areas are shaped by input location - foveal vs. peripheral vision) is evidence for Experience-Driven Development
• Studies on Reading (VWFA) & Expertise (Pokémon studies) show experience-dependent specialization
• Monkeys Deprived of Faces shows that lacking face exposure prevents normal face area development

Evidence for Innate Brain Organization

• Face Processing in Utero is shown in infants orienting toward face-like light patterns
• FFA activates to tactile face recognition in blind individuals, suggesting a pre-existing mechanism for face processing
• Face-selective responses are found as early as 3 months old in Infant fMRI Findings
• Scene-selective and body-selective regions appear less developed
• Suggests that face areas may be more hardwired, while other category areas develop later

Final Notes

• fMRI relies heavily on statistical modeling due to noise and multiple comparisons
• Circular analyses must be avoided through independent data selection
• Face-selective areas may have innate components, that experience refines said specialization

Lecture 8: Neuroeconomics and Neuromarketing

• Neuroeconomics is a multidisciplinary field combining neuroscience, economics, and psychology
• It studies how the brain makes decisions about value, risk, and reward
• Seeks to understand mechanisms driving economic behavior and decision-making
• Neuroeconomics History: Started in the early 1990s
• The field developed and flourished in the early 2000s
• The first textbook was published in 2008
• Neuromarketing is the application of neuroscience to marketing
• Neuromarketing uses brain imaging to understand consumer preferences and is based on the assumption that biology provides a "pure" readout of value
• Neuromarketing bypasses the social/rational filter involved with traditional market research
• A controversial question being is there a "buy button" in the brain?

Brain Systems for Value-Based Decision Making

• The Ventral Medial Prefrontal Cortex (vmPFC) calculates the overall value of stimuli and reflects preference and subjective value of reward
• The Ventral Striatum processes reward and reward prediction error (RPE), receives projections from the dopamine system, and is more active for rewarding stimuli (e.g., sports cars vs. sedans)
• The Anterior Insula computes cost and effort and processes negative emotions
• The Dorsal Anterior Cingulate Cortex (dACC) computes explore/exploit tradeoff and is more active when viewing opposing political candidates
• The Posterior Cingulate Cortex (PCC) extracts value information from memory and tracks long-range probability
• The Hippocampus is involved in memory processes, activates when brand information is present, and helps simulate potential outcomes during deliberation
• The Dorsolateral Prefrontal Cortex (dlPFC) is active when brand information is cued and is involved in cognitive control and decision-making

Decision-Making Process

• General Framework

  Access value and cost priors from memory

  Consider cost to explore

  Calculate overall value

  Execute decision

  Experience rewards/costs

  Update value and cost priors based on experience

How Decisions Are Influenced

• Memory: Prior experiences shape expectations and preferences
• Brand Information: Activates memory-related brain areas and biases judgment
• Deliberation: Mentally simulating options before making a choice
• Erk et al. (2002) - Luxury Cars Study
• This study was funded by Daimler Chrysler research
• The study participants included 12 German men (car enthusiasts)
• Ventral striatum activated more from luxury car stimulation than for sedans or small cars
• vmPFC activated for other car types
• Sports cars indicate higher social rank, therefore more rewarding
• VS is part of the reward system
• vmPFC calculates value through rational means

Political Preference Study

• 20 participants (10 Democrats, 10 Republicans)
• The study had people view images of political candidates (Bush vs. Kerry)
• No brain areas indicated a showed preference for same-party candidates
• The Insula, dlPFC, and dACC lit up more when viewing opposite-party candidates The study had not used a control condition (non-political faces) Possible familiarity confounds No baseline comparison was performed

McClure et al. (2004) - Pepsi Challenge Study

• The study was a Two-phase experiment on soda preference (Coke vs. Pepsi)

• Outside of the scanner: Participants stated preferences Performed blind taste tests (performed at chance level)

• Inside scanner: Associated stimuli with colors Tasted soda and indicated match/no match

• Key findings:

  Without brand information: vmPFC activity correlated with behavioral preference With brand information: Memory-related areas (hippocampus, dlPFC) became active No relationship between vmPFC and preference when brand was shown

• Conclusions:

  Brand information biases brain to use memory-responses (value priors)

  vmPFC reflects subjective interpretation of reward value

• Limitations:

  Possible confound with stimulus types

  Inconsistent results between Coke and Pepsi preferences

Animal Models: Decision-Making and Planning

• Restaurant row studies with mice
• Related to regret: Mice "looked back" after making bad choices More likely to wait and eat faster after experiencing regret
• Neural correlates: When choosing an option: striatum and OFC signal value of current option When experiencing regret: OFC and striatum represent the unchosen option
• Hippocampal role: During deliberation, hippocampus shows alternating activity reflecting possible options Helps simulate outcomes from memory

Critical Evaluation of Neuromarketing

• Limitations of Neuroimaging studies: Small sample sizes Lack of proper control conditions Potential confounds (e.g., familiarity, stimulus properties) Correlational rather than causal evidence
• Ethical Considerations Privacy concerns Potential manipulation of consumer behavior Corporate influence on research
• Scientific Validity Questions regarding neuroimaging reliability in predicting consumer behavior "Brain scam" critique in Nature Neuroscience (2004) The challenge of going from moving lab findings to real-world applications

Lecture 9: Love and Memories

• Concludes discussions on subtraction designs in fMRI studies
• Exam review and Exam 1 preparation will be performed next
• Future topics will include other MRI modalities, plasticity, and technology

Leaning Objectives

• Evaluate if the psychological construct of love be mapped onto brain activation
• Understand adequate controls in experimental design
• Synthesize across studies examining similar brain activity patterns in different contexts (e.g., pain, love, memory)

Parameterizing Love in fMRI Studies

• fMRI requires breaking down cognitive processes into distinct components for contrastive analysis

• Love is a Motivational state that is associated with attachment, caring, and a desire for closeness, as well as a variety of other characteristics. (Reis & Aron, 2008):

  Linked to emotions such as sexual desire, anxiety, jealousy, and grief.

  Behaviorally expressed through proximity-seeking, support, and harmonious contact.

• Monogamy in Nature: Rare but observed in humans, bald eagles, lovebirds, French angelfish, beavers, and prairie voles

• Prairie voles: Vasopressin and oxytocin receptors in reward-associated brain areas

Love in the Human Brain

• Aron et al. (2005) Study:

  Participants: 17 individuals “intensely in love” (10 women, 7 men, relationship duration ~7.4 months).

  Viewing images of a beloved vs. an acquaintance.

• Key Findings:

  Dopaminergic activity in the Ventral Tegmental Area (VTA): Reward signaling

  Caudate nucleus activity: Correlates with self-reported love intensity

Bartels & Zeki (2000) Study

• Similar methodology but with longer relationships (2.5 ± 1.7 years).
• Found activation in caudate and putamen, but not in VTA.
• Suggests long-term love activates different neural circuits

Time in Love and Brain Activity:

Activity in the posterior cingulate cortex (PCC) and medial prefrontal cortex (mPFC) correlates with relationship duration Potential confounds: Task instructions may conflate love with familiarity or memory recall

Love and Pain

• Study by Younger et al. (2010)

• Health Benefits of Love:

  Decreased risk of heart attack, hypertension, and mental health disorders Increased survival rates from surgery and overall lifespan (especially in men)

Behavioral Pain Study - Master et al., 2009

Participants: 28 women in long-term relationships (>6 months). Method: Thermal pain applied to forearm under different conditions (partner vs. stranger vs. object). Result: Holding or viewing a partner reduced perceived pain

FMRI Study on Love & Pain:

Participants: 15 individuals in relationships (<9 months). Viewing partner's image vs. acquaintance during pain stimulation. Results: Increased activity in Nucleus Accumbens (NAcc) and Caudate. Activity correlated with pain relief, suggesting a role of the reward system in pain modulation Issues: Eye movements can distort fMRI signals, leading to potential artifacts

Memory & Love (Bainbridge & Baker, 2022)

• Previous studies confounded love with memory strength, temporal distance, emotion, and familiarity
• Study used “One Video per Day dataset” (300 personal vs. 300 other-user videos) Measured memory strength, emotional intensity, and temporal/physical distance
• Key Findings:

Familiarity preference: Large portions of the brain activated for personal over other-user videos

PCC activity correlated with memory strength, similar to prior studies linking PCC to love duration A study showed that Emotion modulated activity in limbic and visual areas (e.g., LOTC, Amygdala, NAcc)

Implications For Studying Complex Cognitive Processes

• Psychological constructs (e.g., love, pain, decision-making) are not physical entities but useful models for understanding behavior

• Inside-Out Approach (Buszaki, 2019):

  Instead of mapping psychology onto the brain, researchers should let brain activity define meaningful categories. Utilize large datasets to extract features that the brain itself prioritizes

Explore the fundamentals of fMRI, contrasting it with MRI. Learn how fMRI measures brain activity using the BOLD signal and its advantages, including non-invasiveness and wide availability. Understand its role in cognitive and neuroscience research.