Brain Imaging Methods: What MRI, fMRI, PET, and EEG Measure
Author: Ukrainian Psychological Hub · Published: October 12, 2026 · Editorial Policy
Brain imaging methods allow researchers and clinicians to investigate different properties of the living brain. Structural magnetic resonance imaging (MRI) depicts anatomical and tissue characteristics. Functional MRI (fMRI) measures changes in magnetic resonance signals related to blood oxygenation and other physiological contrasts. Positron emission tomography (PET) detects the distribution of radioactive tracers chosen to probe particular biochemical processes. Electroencephalography (EEG) records voltage differences at the scalp produced by coordinated electrical activity in neural tissue. These signals arise from different physical processes, so they answer different questions.
The central principle of neuroimaging is that an image is a measurement transformed into a representation. Color on an fMRI map is usually a statistical display; color on a PET image shows estimated tracer distribution or a derived parameter; MRI contrasts depict signal differences influenced by tissue properties; EEG traces show voltage changes over time. The images are powerful scientific evidence when linked to a defined experiment and suitable interpretation. They are not photographs of emotions, memories, intentions, or individual neurons.
The distinction between measurement and interpretation is especially important in psychology. A person may perform a memory task, experience anxiety, or respond to a social signal while several neural and bodily systems change together. A brain recording supplies evidence about one part of that process. Behavioral measures, experimental controls, longitudinal data, and theoretical models determine what conclusions the recording supports. The NIH's MRI overview and nuclear medicine overview describe the underlying technologies.
MRI, fMRI, PET, and EEG: the essential comparison
Structural MRI: anatomy and tissue contrast
MRI detects radiofrequency signals from nuclear spins, predominantly hydrogen in body water and fat, in a strong magnetic field. Different sequences weight relaxation, proton density, diffusion, magnetic susceptibility, or other tissue-dependent properties. The main outputs are anatomical volumes and quantitative or semiquantitative tissue maps. Conventional structural MRI excels at locating lesions, comparing gross anatomy, and measuring features such as cortical thickness or regional volume. Anatomical shape is not a direct measure of cognitive performance, neuronal firing, or a diagnosis of a psychiatric condition.
Functional MRI: hemodynamic changes associated with neural processes
Most cognitive fMRI uses blood-oxygen-level-dependent (BOLD) contrast, a magnetic resonance signal influenced by changes in deoxygenated hemoglobin, blood flow, blood volume, and oxygen consumption. BOLD responds to neurovascular coupling between neural and vascular activity. Maps identify locations whose signals differ under specified conditions or fluctuate together over time. Spatial sampling is commonly in millimeters; the physiology develops over seconds even when acquisition volumes are collected relatively quickly. The BOLD signal is an indirect, context-dependent index of aspects of neural activity, as explained by Logothetis (2008). For a dedicated explanation of BOLD physiology and the limits of fMRI interpretation, see Functional MRI: How fMRI Measures Brain Activity and Its Limits.
PET: tracer-specific metabolism or molecular targets
PET records pairs of photons resulting from positron annihilation after a radiotracer has been administered. The tracer determines what biological question can be addressed. Fluorodeoxyglucose (FDG) studies glucose uptake and phosphorylation-related trapping in tissue; other tracers target selected receptors, transporters, enzymes, proteins, or other processes. PET therefore measures tracer behavior, not a generic unit of brain activity. It offers molecular specificity that BOLD fMRI and routine EEG usually do not, at the cost of ionizing radiation, tracer-dependent procedures, and slower effective time scales.
EEG: the timing of large-scale neural electrical activity
EEG measures differences in electrical potential between electrodes, generally attached to the scalp. Its recorded signals are strongly influenced by the summed extracellular fields generated by synchronized transmembrane currents in populations of neurons; cortical pyramidal cells are prominent contributors. EEG can resolve changes on millisecond time scales and has long been used to examine sensory responses, attention, sleep, and seizure-related activity. Spatial localization requires modeling the conductivity of head tissues and remains uncertain, especially for deep or complex sources. The American Epilepsy Society EEG text provides the biophysical explanation.
Magnetoencephalography (MEG) provides another perspective on fast neural activity by recording weak magnetic fields associated with coordinated neuronal currents. It complements EEG and can contribute to selected clinical functional mapping and epilepsy evaluations; source maps remain model-based estimates. For the measurement physics, practical procedure, temporal precision, and interpretive limits, see our detailed explanation of magnetoencephalography (MEG).
What does resolution mean in brain imaging?
Spatial resolution refers to the ability to distinguish signals originating from nearby locations. Temporal resolution describes how precisely a method can distinguish changes in time. Molecular specificity describes the extent to which a signal is selective for a biological target. Sensitivity describes the ability to detect the signal of interest against noise. All four matter, and no method simultaneously maximizes them in every biological context.
A typical structural MRI voxel may represent approximately a millimeter-scale volume, depending on the sequence and scanner, and specialized acquisitions can resolve finer details. A typical fMRI voxel may sample a few millimeters, whereas ultra-high-field systems can acquire submillimeter-scale information in selected protocols. Yet nominal voxel size is not identical to independently resolved neural populations: vascular spread, blurring, distortion, noise, and analysis choices remain important. A 2026 Neuron primer on ultra-high-field fMRI discusses both the promise of cortical-layer imaging and the substantial interpretation challenges.
EEG samples voltage at high frequency and can align averaged brain responses to stimulus onsets with millisecond precision. Its scalp electrodes receive mixtures of signals from distributed generators; a large number of electrodes improves the data available for source modeling without eliminating the underlying inverse problem. PET scanners can localize tracer distribution at millimeter scales under suitable conditions, but estimated biological changes often require accumulation and kinetic modeling over minutes. These are deliberately contextual descriptions rather than universal numerical rankings: scanner generation, tracer chemistry, anatomy, acquisition, and analysis determine actual performance.
The temporal resolution of the instrument must also be separated from the temporal specificity of the underlying biological phenomenon. A rapid fMRI sampling interval does not make the delayed hemodynamic response a millisecond neural recording; a fast PET acquisition frame may contain too little tracer signal to estimate meaningful molecular changes; a high EEG sampling frequency cannot recover invisible deep sources. Earlier comparative methodology reviews made this point explicitly, including Aine's evaluation of MRI, fMRI and PET.
How MRI forms a brain image
Hydrogen nuclei possess magnetic moments. Inside an MRI scanner, the net magnetization of many nuclei is influenced by the main magnetic field. Radiofrequency pulses alter that magnetization; receiver coils detect signals as it evolves. Applied gradients encode spatial position. Reconstruction algorithms transform the sampled signals into images, with image contrast shaped by pulse sequence, relaxation behavior, motion, and field properties. Medical images are normally reviewed as cross-sectional slices and three-dimensional volumes.
T1-weighted imaging commonly offers useful gray–white matter contrast for measuring anatomy and visualizing structural changes. T2-weighted and FLAIR imaging emphasize other tissue differences and are valuable in identifying various lesions or abnormal fluid-related signals. Susceptibility-sensitive sequences respond to local magnetic field perturbations, including effects associated with blood products or iron-containing tissue. Quantitative MRI attempts to estimate specified tissue-related parameters under modeled conditions. A difference in image intensity is not itself a count of neurons, synapses, or particular neurotransmitters.
Diffusion MRI extends the family by measuring how water movement is constrained by tissue microstructure. Diffusion tensor and more advanced models estimate directional properties of water diffusion; tractography then uses mathematical tracking rules to infer plausible trajectories through white matter. Streamlines are modeled pathways, not directly observed individual axons. The 2018 validation challenge documented continuing anatomical limitations and the importance of independent validation. This branch of MRI is especially relevant when a study claims to have mapped brain connections.
Structural MRI has clear clinical uses in evaluating many neurological conditions, including tumors, inflammation, vascular injury, and anatomical abnormalities, under appropriate indications. Its strong tissue contrast is a major reason it is widely used in the brain. MRI involves a powerful magnet and radiofrequency energy rather than ionizing radiation. Actual safety assessment depends on implant compatibility, possible metal fragments, pregnancy context, kidney function when certain contrast agents are contemplated, and a person's ability to tolerate the procedure. RadiologyInfo's brain MRI guide outlines the practical screening process.
How fMRI measures changes during cognition and behavior
Functional MRI is a family of MRI methods rather than a single measure. BOLD fMRI is dominant in cognitive neuroscience, while arterial spin labeling estimates perfusion using magnetically labeled blood water and some experiments use other contrasts. In a typical task-fMRI experiment, investigators specify events, conditions, behavior, and an analysis plan. The scanner collects a time series of images; mathematical models estimate which regions show signal changes related to the events after allowing for the expected hemodynamic response.
The BOLD signal and neurovascular coupling
Neural signaling changes local energy demand and the behavior of vascular and supporting cells. The resulting changes in blood flow, blood volume, and oxygen extraction alter the local concentration and distribution of deoxyhemoglobin, which affects an MRI signal sensitive to magnetic susceptibility. Under common task conditions, the local BOLD signal often increases as oxygenated blood supply exceeds oxygen use. But different mixtures of vascular and metabolic responses can produce different time courses, and the relationship is altered by vascular health, age, medication, carbon dioxide, and other physiological states.
Accordingly, the phrase 'this brain area lit up' means that an imaging contrast met a defined statistical criterion relative to a baseline or contrast. A positive BOLD response does not prove that all local neurons increased their spike rate, that inhibitory signaling was absent, or that a discrete mental faculty occupies the colored region. Negative BOLD responses likewise require physiological and experimental interpretation. Logothetis's review remains a foundational account of what neurovascular signals allow researchers to infer.
Task-based contrasts, resting-state connectivity, and design
Task fMRI tests differences between experimental conditions. For example, the question may be whether brain signals during successful memory encoding differ from those during an appropriate comparison task. The estimate depends on the chosen baseline, event timing, behavioral performance, head motion, attention, and model specification. Properly controlled contrasts help adjudicate hypotheses; uncontrolled differences can mix the target process with difficulty, sensory input, movement, fatigue, or arousal.
Resting-state fMRI asks how BOLD fluctuations covary across regions without a prescribed task. Correlated fluctuations can support estimates of functional connectivity and reveal reproducible features of large-scale network organization. They do not demonstrate direct axonal connections or that one region caused another to activate. Physiological fluctuations, head movement, scanner drift, and preprocessing alter correlations. Murphy, Birn, and Bandettini's review of resting-state confounds and Power et al.'s motion analysis explain why the controls are central to the result.
Statistical maps and multiple comparisons
A functional brain image typically includes thousands of voxels or surface locations. Testing many locations raises the probability of false-positive discoveries unless multiplicity is handled appropriately. Researchers use prespecified hypotheses, correction methods, permutation procedures, effect-size estimates, cross-validation, and independent replication to support inference. It matters whether a result refers to a voxel, a cluster, a region selected after looking at the same data, or a model tested on previously unseen participants.
A landmark analysis by Eklund, Nichols, and Knutsson (2016) demonstrated inflated false-positive rates for some historically common cluster-inference procedures under the tested assumptions and settings. The relevant lesson is precise statistical validation and updated practices, not a blanket claim that every past or present fMRI finding is false. Different procedures, quality checks, and study designs have different operating characteristics.
How PET measures brain chemistry and metabolism
PET uses a radioactive isotope attached to a molecule with a defined biological distribution or interaction. After administration, the radiotracer emits positrons as the isotope decays. Each positron eventually interacts with an electron, generating two photons traveling in nearly opposite directions. Detectors register coincident photons, and reconstruction procedures estimate the distribution of detected radioactivity. Motion correction, attenuation correction, scanner calibration, and kinetic modeling influence the resulting image.
FDG PET and energy metabolism
FDG is a radiolabeled glucose analog transported into cells and phosphorylated. Its accumulation over the relevant period is used to estimate regional patterns of glucose use, subject to the tracer's kinetics and other assumptions. Cellular energy demand includes neuronal signaling, glial activity, maintenance processes, and multiple metabolic pathways. Consequently, FDG uptake should be interpreted as a tracer-derived metabolic measure rather than a direct neuron count or a photograph of thought.
Receptor imaging and the meaning of binding
Specialized PET ligands can estimate properties associated with the availability of certain receptors or transporters, or the deposition of targeted proteins, provided the tracer and measurement model are appropriate. For receptor studies, a parameter such as binding potential depends on multiple factors, including receptor availability and ligand affinity, and may be influenced by competition with endogenous compounds. Different tracers have distinct properties, so a 'dopamine PET scan' is scientifically incomplete without the tracer, endpoint, and kinetic model.
PET measurements also encounter partial-volume effects: signals from small structures mix with neighboring tissues because finite spatial resolution and reconstruction spread the signal. Brain atrophy can complicate apparent differences between groups. Kanel et al. (2023) review these issues for molecular brain PET. A between-group difference in a tracer-derived quantity neither proves a simple deficiency in a neurotransmitter nor establishes a cause of a person's behavior.
PET is used in research to study receptor systems, pharmacological occupancy, metabolism, and selected neurodegenerative biomarkers. Some radiotracers have specific authorized clinical uses; clinical eligibility and diagnostic meaning depend on the exact indication, tracer, regulatory status, and accompanying clinical evaluation. Because radioactive material is administered, radiation dose and possible pregnancy or breastfeeding implications require protocol-specific review. The authoritative physical and safety overview is provided by NIH/NIBIB nuclear medicine.
How EEG records electrical brain signals
EEG electrodes measure voltage differences on the scalp relative to selected references. Brain cells continuously exchange ions across their membranes, generating microscopic currents and associated electric fields. When appropriate neuronal populations have aligned geometry and sufficiently coordinated activity, the combined field can be detected outside the skull. Cortical postsynaptic and other transmembrane currents contribute substantially to conventional scalp EEG, with cortical pyramidal-cell organization being particularly important. EEG is therefore a population-level electrical recording, not a direct readout of a single neuron's action potentials.
Event-related potentials and oscillations
When an experiment presents repeated stimuli or responses, researchers can average EEG segments aligned to each event to estimate event-related potentials (ERPs). Components of an ERP reflect processes unfolding within a particular paradigm, including sensory discrimination, attention, prediction, and decision-related activity. Component names describe characteristic timing, polarity, or experimental relationships; they do not establish that a waveform is a one-to-one psychological trait or clinical diagnosis.
Spectral analysis examines how electrical signal power and coordination vary across frequencies. Conventional labels such as delta, theta, alpha, beta, and gamma summarize frequency ranges, with conventions differing across laboratories and populations. Frequency-band power can change with sleep, vigilance, sensory engagement, movement, and illness. There is no universal mapping in which one 'brain wave' directly equals intelligence, creativity, calmness, a specific emotion, or psychological wellness.
Source localization and the inverse problem
A scalp voltage pattern can be produced by more than one configuration of neural generators. This is the EEG inverse problem. Source-imaging algorithms incorporate physical head models, electrode positions, and assumptions or regularization to estimate where contributing activity may have originated. Those estimates can be improved by individual MRI anatomy and dense electrode arrays, but they remain model dependent. Skull conductivity, electrode contact, muscle activity, eye movements, and other artifacts introduce additional uncertainty. The American Epilepsy Society's biophysical account and a 2026 review of EEG generators explain the mechanisms.
EEG in clinical practice
EEG is an established component of epilepsy evaluation and selected other clinical investigations. Clinicians interpret waveforms, event descriptions, history, examination, and sometimes video recordings together. A normal routine EEG does not exclude epilepsy; a nonspecific rhythm abnormality does not by itself diagnose ADHD, depression, anxiety, or another psychological disorder. The Epilepsy Foundation's EEG guidance explains routine, prolonged, and ambulatory recording, and its normal-EEG explanation clarifies sensitivity limits.
What can and cannot be inferred from a brain image?
Association, experimental contrast, and causal explanation
An anatomical correlation between regional volume and a behavior suggests a relationship measured in a specified sample. A functional contrast shows that a signal differs across experimental conditions. A longitudinal scan shows change over time. None automatically identifies why the change occurred. To test causation, researchers strengthen designs with randomization when appropriate, interventions, prospective data, lesion evidence, stimulation experiments, genetically informative comparisons, or other converging approaches. Every method has its own inferential constraints.
Brain imaging also interacts with behavior in both directions. Learning and behavior can alter neural processing, and neural conditions can influence cognition and action. Sleep, movement, medications, stress, vascular factors, age, socioeconomic exposures, and scanner protocol can affect measurements. A credible neural explanation therefore incorporates mechanisms together with the psychological and environmental conditions in which a person acts.
Reverse inference and mental-state claims
Reverse inference means inferring a particular mental process because a brain region or network displayed a characteristic signal. The inference is stronger when there is good evidence that the pattern selectively predicts that process in the relevant task and population; it is weak when the same pattern appears during many unrelated processes. Poldrack (2006) formulated the caution, while Hutzler (2014) examined circumstances in which conditionally specified reverse inference can be informative.
A study reporting an amygdala response to emotional faces cannot simply conclude that its participants 'felt fear.' Frontal activity during a difficult task does not alone identify self-control, deception, or conscious intention. Network patterns associated with verbal information cannot be assumed to decode private sentences under unconstrained daily-life conditions. Even when a trained classifier predicts a task label, prediction must be distinguished from an independently validated explanation of an individual's mental state.
Why attractive colored maps can be misleading
Most brain images are displayed after acquisition, denoising, reconstruction, spatial normalization, smoothing, thresholding, modeling, and visualization choices. A red or blue voxel may represent the sign and magnitude of a modeled contrast, a standardized statistic, a tracer concentration, or another numerical quantity. Color scales are selected by software or investigators. The reader should always ask: What was measured? Compared with what baseline? What correction was applied? How large was the effect? How consistent was it across participants? How was the hypothesis tested on independent data?
Reliability, replication, and individual brain biomarkers
Group-average research findings and individual prediction are distinct scientific goals. A measure may reliably show an average task response yet rank the same individuals inconsistently across repeated sessions. Low test–retest reliability limits its usefulness as an individual biomarker even when the group's activation map appears orderly. In a meta-analysis of task-fMRI reliability, Elliott and colleagues summarized 90 experiments and reported a mean intraclass correlation of approximately 0.40 across included estimates, a limitation for many individual-difference applications. Reliability varies by protocol and can improve through better design and repeated measurement; it must be demonstrated for the proposed use.
A clinically useful biomarker must meet more than one threshold: analytic validity, reliability, meaningful sensitivity and specificity, appropriate reference populations, prospective external validation, clinical utility, and acceptable benefit–risk balance. MRI is clinically established for many structural and neurological indications; EEG has validated applications in seizure assessment; some PET tracers support specific neurological or oncological questions. Their existence does not make research correlations into diagnostic tests for depression, ADHD, autism, anxiety, or personality.
This distinction is illustrated in the English Hub's existing clinical review Depression and the Brain: What Neuroscience Actually Shows, where imaging findings are discussed as population associations with substantial person-to-person overlap. The clinical definitions and treatment questions remain with their condition-specific articles; this article owns the comparative imaging-method question.
Combining methods: what multimodal neuroscience adds
Multimodal studies use complementary measurements to constrain interpretation. Structural MRI can supply anatomy for locating an EEG source estimate. EEG recorded alongside fMRI may relate rapid electrical fluctuations to slower BOLD changes. PET/MRI can combine a tracer-sensitive measurement with anatomical and functional MRI information. Behavioral accuracy, eye tracking, physiological monitoring, and carefully designed task manipulations can further clarify the human processes involved.
Combining instruments does not magically turn correlations into causal evidence or remove measurement error. Simultaneous EEG-fMRI requires accounting for MRI-induced EEG artifacts, physiological noise, and synchronization. PET-MRI requires accurate registration and recognition that tracer behavior and BOLD reflect different properties. Interpretation becomes stronger when the methods produce mutually constraining results under prespecified models and when independent samples replicate the relationships.
How to choose the right method for a neuroscience question
For gross anatomy, focal lesions, tissue characteristics, or macrostructural change, conventional and specialized MRI sequences are generally the first family of measures to evaluate. If the question concerns where an experimentally defined cognitive task is associated with changes in hemodynamics, task fMRI may be appropriate. If the question concerns fluctuations among distributed systems at rest, resting-state fMRI offers a statistical approach to covariation that needs careful physiological controls.
If the question concerns rapid brain responses to a sound, an error signal, a visual event, sleep stage, or possible epileptiform activity, EEG is usually the more appropriate member of this comparison for timing. If the question specifically concerns glucose-tracer uptake, receptor availability, transporters, or a tracer-defined protein target, PET may be necessary. When a study needs both temporal and anatomical detail, an integrated design may offer the best scientific leverage.
For diagnosis, the right test depends on symptoms, history, examination, established clinical indications, local safety protocols, and the question being asked. The most sophisticated scanner is not automatically the most informative procedure for a particular person. Researchers must also weigh participant burden, cost, recruitment feasibility, data quality, availability of expertise, repeatability, and the population to which results will be generalized.
Research limitations that apply across technologies
Sampling, confounders, and demographic generalization
Neuroimaging findings can be sensitive to age, developmental stage, sex-related biology, neurological and medical comorbidity, medications, sleep, motion, vascular physiology, handedness, accessibility, and scanner differences. Samples drawn from university populations or a small number of hospitals may not represent the population to which an application is advertised. Multisite validation and transparent reporting matter, especially for clinical translation and machine-learning classification.
Preprocessing, thresholds, and analytical flexibility
Reconstruction and preprocessing are part of measurement rather than cosmetic steps. Choices in spatial smoothing, motion correction, statistical thresholds, region definitions, EEG reference selection, frequency filtering, PET kinetic models, and partial-volume correction can change outputs. Preregistration where appropriate, code and data sharing within privacy limits, pipeline sensitivity analysis, and independent replication help distinguish robust findings from analysis-dependent results.
Machine learning and neural decoding
Algorithms can extract patterns from MRI, fMRI, PET, or EEG and predict labels under controlled experimental conditions. Strong demonstrations require genuinely independent evaluation data, protection against information leakage, relevant baselines, and tests across scanners and populations. Decoding a stimulus class or a constrained language task is a different claim from recovering unrestricted private thoughts. Performance statistics must be attached to the studied task, participants, recording conditions, and model. The detailed question of 'mind reading' belongs to EN-30 N50 in the planned knowledge graph, while the measurement basis belongs here.
Rapid developments through 2026
Ultra-high-field fMRI, improved acquisition hardware, personalized repeated scanning, more sensitive PET instrumentation, new molecular tracers, and advanced electrophysiological source models continue to extend the scope of human neuroscience. The 2026 ultra-high-field fMRI primer explicitly emphasizes that higher spatial resolution creates additional interpretive demands. New techniques should be evaluated using their actual validation studies, rather than assuming that better-looking images have solved the physical or biological inference problem.
Popular brain-imaging myths, examined
Myth: fMRI photographs active neurons. The dominant BOLD method measures a physiological contrast associated with vascular and metabolic responses, interpreted through models of neurovascular coupling. It does not directly record individual neuronal spikes.
Myth: PET measures the amount of dopamine or serotonin everywhere in the brain. PET detects the distribution and kinetics of a selected radiotracer. A receptor-ligand binding parameter may reflect availability and competition under model assumptions; it is not a comprehensive readout of neurotransmitter concentration or mood.
Myth: MRI proves that a psychological experience permanently changes the number of brain cells. Routine MRI measures tissue-related signals and macroanatomical estimates; observed differences in volume or thickness have multiple potential biological and analytic explanations.
Myth: EEG measures thoughts directly and identifies precise deep-brain locations. Scalp EEG measures spatial mixtures of electrical fields with very good timing. Localizing generators and mapping them to subjective contents requires additional assumptions and evidence.
Myth: a strong group-average imaging finding can automatically diagnose one person. Population differences, individual classification, test–retest reliability, and clinical benefit require separate validation. Many measures overlap substantially between groups.
Myth: one colored activation map establishes a brain's emotional center. Regions participate in distributed, multifunctional circuits, and brain signals are influenced by tasks, prior learning, the body, and context. Interpretations gain specificity through experimental comparisons and converging evidence.
Frequently asked questions
Which test actually shows the brain's structure?
Structural MRI is the principal method among these four for detailed anatomical and tissue-contrast images. Specialized MRI sequences can also probe diffusion and other physical tissue properties. PET can be anatomically aligned with MRI or CT but is primarily tracer-based.
Which brain test measures activity most directly?
Scalp EEG records electrical potentials arising from neuronal population activity, making it temporally close to rapid neural processes. It is still a spatial mixture of electrical fields. BOLD fMRI and most PET applications infer relevant processes from hemodynamic or tracer-related signals.
Is fMRI the same as a normal MRI?
Both use magnetic resonance physics. A conventional anatomical MRI emphasizes structural tissue contrast, whereas most cognitive fMRI involves repeated scanning and models of blood-oxygenation-related signal changes during tasks or at rest. The appropriate protocol depends on the question.
Why is fMRI slower than EEG?
The BOLD response is governed by vascular and metabolic dynamics that typically unfold over seconds. EEG records electrical fields that can change within milliseconds. Faster MRI acquisition does not eliminate the slower underlying hemodynamic response.
Does a brighter fMRI image mean more thinking or intelligence?
No. Brightness and color are chosen visualizations of a specific signal or statistic. Larger signals do not consistently mean more effective cognition, greater intelligence, greater effort, or stronger subjective experience.
Can PET show dopamine or serotonin?
Selected radiotracers can be used to investigate certain receptors, transporters, enzymes, or other molecular processes relevant to these neurotransmitter systems. Interpretation is tracer- and model-specific; routine PET is not a direct total-dopamine or total-serotonin meter.
Does EEG read individual neurons?
Routine scalp EEG mainly reflects summed activity of populations of neurons whose electric fields reach electrodes through intervening tissues. It cannot ordinarily separate action potentials from single cells. Intracranial recordings and single-unit methods address different, often invasive measurements.
Are EEG brain-wave bands reliable signs of a mental state?
Frequency-band measures can be useful in specified experiments or clinical contexts. Their meaning depends on region, reference, artifact control, age, task, sleep–wake state, and other physiological variables. A single band is not a validated universal label for an emotion or personality trait.
Can brain imaging prove that two brain regions are connected?
Diffusion MRI can estimate possible anatomical fiber pathways, and fMRI can estimate functional covariation. Both are model-based: tractography can produce missed and false-positive pathways, and functional connectivity does not establish direct anatomy or causation. Other anatomical and experimental methods may strengthen the claim.
Can a brain scan diagnose ADHD, depression, or anxiety?
Current routine diagnosis of these conditions is based on appropriate clinical criteria and assessment, not an individual MRI, fMRI, PET, or EEG signature. Research may identify group associations or promising candidate biomarkers, but those findings need independent clinical validation before becoming diagnostic tests. Clinically ordered scans may be useful to evaluate other conditions when justified.
Are MRI, fMRI, PET, and EEG safe?
Risk depends on the protocol and individual circumstances. MRI and BOLD fMRI avoid ionizing radiation but require metal and implant screening, and contrast procedures have additional considerations. Routine scalp EEG is noninvasive. PET involves administration of a radioactive tracer and a medical decision about radiation exposure. The clinician or research team should describe the protocol's actual risks and benefits.
Which is better for studying memory, emotions, or attention?
The method should follow the hypothesis. EEG can characterize fast processing time courses; fMRI can examine spatial patterns of hemodynamic responses; MRI can assess anatomical features; PET can probe selected molecular targets. Psychological tasks and behavior remain essential for determining what 'memory,' 'emotion,' or 'attention' means in the experiment.
Can two methods be used together?
Yes. EEG-fMRI, PET-MRI, and structural MRI-guided EEG source analysis can add complementary information. Their value depends on data quality, synchronized design, suitable analysis, and claims justified by the joint measurements.
Related Articles
For an in-depth, tracer-specific explanation, read PET Brain Imaging: Metabolism, Receptors, and Research Applications.
For a clinical example of how to interpret research neuroimaging findings without turning population-level differences into an individual diagnosis, read Depression and the Brain: What Neuroscience Actually Shows. Other EN-30 methods articles—including the dedicated fMRI, EEG, PET, MEG, and neural decoding articles—will be linked when their own English canonical pages have been published.
The limits of inferring individual neuron properties from fMRI, EEG, and regional activation overlap become particularly clear in Mirror Neurons: Discovery, Evidence, and Common Misconceptions.
For the research design that relates structural damage to functional necessity, see brain lesion studies.
References
Aine, C. J. (1995). A conceptual overview and critique of functional neuroimaging techniques in humans: I. MRI/fMRI and PET. Critical Reviews in Neurobiology, 9, 229–309. PubMed 8581985.
Beniczky, S., & Scherg, M. (2026). Scalp EEG signals: Generators and dipoles. Journal of Clinical Neurophysiology. DOI: 10.1097/WNP.0000000000001226.
Britton, J. W., Frey, L. C., Hopp, J. L., et al. (2016). The scientific basis of EEG: Neurophysiology of EEG generation in the brain. In Electroencephalography (EEG): An Introductory Text and Atlas. American Epilepsy Society. NCBI Bookshelf.
Eklund, A., Nichols, T. E., & Knutsson, H. (2016). Cluster failure: Why fMRI inferences for spatial extent have inflated false-positive rates. Proceedings of the National Academy of Sciences, 113, 7900–7905. DOI: 10.1073/pnas.1602413113.
Elliott, M. L., Knodt, A. R., Ireland, D., et al. (2020). What is the test–retest reliability of common task-functional MRI measures? New empirical evidence and a meta-analysis. Psychological Science, 31, 792–806. DOI: 10.1177/0956797620916786.
Epilepsy Foundation. (2025). Electroencephalography (EEG). Official clinical information.
Epilepsy Foundation. What if the EEG is normal? Official clinical information.
Hutzler, F. (2014). Reverse inference is not a fallacy per se: Cognitive processes can be inferred from functional imaging data. NeuroImage, 84, 1061–1069. DOI: 10.1016/j.neuroimage.2012.12.075.
Kanel, P., Carli, G., Vangel, M., Roytman, S., & Bohnen, N. I. (2023). Challenges and innovations in brain PET analysis of neurodegenerative disorders: A mini-review on partial volume effects, small brain region studies, and reference region selection. PubMed Central.
Logothetis, N. K. (2008). What we can do and what we cannot do with fMRI. Nature, 453, 869–878. DOI: 10.1038/nature06976.
Murphy, K., Birn, R. M., & Bandettini, P. A. (2013). Resting-state fMRI confounds and cleanup. NeuroImage. PubMed 23571418.
National Institute of Biomedical Imaging and Bioengineering (NIBIB). Magnetic resonance imaging (MRI). National Institutes of Health. Official source.
National Institute of Biomedical Imaging and Bioengineering (NIBIB). Nuclear medicine. National Institutes of Health. Official source.
Poldrack, R. A. (2006). Can cognitive processes be inferred from neuroimaging data? Trends in Cognitive Sciences, 10, 59–63. DOI: 10.1016/j.tics.2005.12.004.
Power, J. D., Mitra, A., Laumann, T. O., et al. (2014). Methods to detect, characterize, and remove motion artifact in resting state fMRI. NeuroImage, 84, 320–341. DOI: 10.1016/j.neuroimage.2013.08.048.
RadiologyInfo.org (American College of Radiology/Radiological Society of North America). Brain MRI. Official patient information.
Research primer: fMRI at ultra-high field: From acquisition to interpretation. (2026). Neuron, 114, 3133–3155. DOI: 10.1016/j.neuron.2026.04.025.
Schilling, K. G., Gao, Y., Janve, V., et al. (2019). Limits to anatomical accuracy of diffusion tractography using modern approaches. NeuroImage. PubMed 30317017.