Why Epilepsy Tests Sometimes Disagree
What does it mean when the EEG, MRI, PET or other investigations seem to point in different directions?
Epilepsy investigation can sometimes look confusing.
One test may suggest seizures are coming from the left side of the brain.
Another may appear bilateral.
MRI may show an abnormality somewhere else.
PET might reveal a larger area of altered metabolism.
Neuropsychological testing may suggest dysfunction in a somewhat different network.
And occasionally a person's symptoms seem not to fit neatly with any of them.
This does not automatically mean one of the tests is wrong.
Different investigations measure different aspects of the brain:
structure
electrical activity
metabolism
blood flow
magnetic fields
cognition
behaviour
and brain function.
They also examine the brain:
at different times
under different conditions
and with different spatial and temporal limitations.
Modern epilepsy assessment therefore relies on multimodal evidence rather than expecting one investigation to provide the entire answer.
Presurgical epilepsy evaluation is specifically designed to combine several sources of information because no individual modality provides absolute information about the complete epileptogenic network.
What does “concordant” mean?
Clinicians often use the word:
concordant
when several pieces of evidence support the same overall explanation.
For example:
the seizure symptoms suggest a left temporal origin;
video EEG records seizures predominantly beginning over the left temporal region;
MRI shows a left temporal structural abnormality;
and PET shows compatible left temporal hypometabolism.
These findings are concordant.
They do not necessarily identify exactly the same millimetre of brain tissue.
They support the same broader hypothesis.
What does “discordant” mean?
Discordant means important results do not appear to support the same localisation or interpretation.
For example:
MRI suggests a right temporal lesion;
but several recorded seizures appear to begin electrically on the left.
Or:
scalp EEG appears bilateral;
but seizure symptoms and functional imaging strongly suggest one focal region.
Discordance does not automatically mean:
“the tests are useless.”
It means the team needs to understand why the results differ before making an irreversible treatment decision.
Concordance is especially important before surgery
Epilepsy surgery involves treating brain tissue.
Doctors therefore need strong evidence that:
the correct seizure-generating network has been identified
and the proposed treatment can be performed safely.
Research and clinical practice consistently show the value of combining:
seizure history and semiology
video EEG
structural MRI
neuropsychology
functional imaging
and, when necessary, intracranial recordings.
A recent 2026 systematic review specifically examining mesial temporal lobe epilepsy with hippocampal sclerosis found that concordance between MRI, unilateral EEG findings and PET was associated with better postoperative seizure outcomes. That result applies to the population studied rather than every epilepsy type, but it illustrates why multimodal agreement matters.
Concordance does not mean every test must agree perfectly
This is important.
The brain is not divided into isolated boxes where every investigation should identify exactly the same boundary.
An epilepsy network may include:
the area where seizure activity begins
regions recruited immediately afterwards
regions repeatedly affected between seizures
areas with altered metabolism
and networks affected cognitively by years of epilepsy.
Different investigations can therefore identify different parts of the same network.
That can still be clinically coherent.
Why can scalp EEG and MRI disagree?
MRI examines brain structure.
EEG examines electrical activity.
A structural abnormality does not automatically prove that seizures start there.
Likewise, electrical seizure activity may occur in tissue that appears structurally normal on MRI.
The two tests therefore answer different questions.
NICE explicitly describes neuroimaging as a way to investigate structural causes rather than as a standalone test for whether epilepsy exists.
An MRI abnormality may not be the seizure source
Suppose MRI reveals a small abnormality.
It might be:
the actual epileptogenic lesion
part of the seizure network
an old unrelated abnormality
or an incidental finding.
Doctors therefore ask:
Does the seizure history fit it?
Does EEG support it?
Do functional investigations support it?
Do recorded seizures begin in or near that region?
MRI alone usually cannot answer all of those questions.
The opposite can happen: epilepsy with apparently normal MRI
A normal MRI does not mean focal epilepsy is impossible.
Some epileptogenic abnormalities can be:
extremely subtle
below conventional imaging resolution
difficult to recognise
located within complex cortex
or not associated with a structural lesion visible on current MRI.
NICE recommends specialist neuroradiology review when seizures continue despite treatment and the diagnosis remains unclear, and consideration of repeat MRI when surgery is being considered.
“MRI-negative” can change after specialist review
The term MRI-negative epilepsy means no relevant structural lesion has been identified on the available MRI.
It does not necessarily mean the brain is structurally completely normal.
A 2025 systematic review found wide variation in how MRI-negative epilepsy was defined and reported. It also found that specialist imaging techniques and MRI post-processing could identify abnormalities in some people originally considered MRI-negative.
This is one reason an epilepsy-surgery centre may:
review the original images again
repeat MRI using a specialist epilepsy protocol
or use additional image-processing methods.
Why can scalp EEG appear bilateral when seizures are actually focal?
Scalp EEG records electrical activity after it has travelled from the brain to electrodes on the scalp.
Several things can make localisation difficult.
A seizure may:
begin deep inside the brain
begin in a very small cortical area
spread extremely rapidly
or become visible on scalp EEG only after both sides of the brain are involved.
The recording may therefore appear:
widespread
bilateral
poorly localised
or even initially normal.
That does not automatically mean the seizure began simultaneously across both hemispheres.
Deep seizures can be particularly difficult for scalp EEG
Some brain regions are relatively distant from scalp electrodes.
Examples include parts of:
the hippocampus
amygdala
insula
and deep cortical folds.
A small deep focal seizure may generate too little scalp signal to be clearly recognised until it spreads.
This is one reason selected people eventually require intracranial EEG or SEEG.
Related Information Hub page:
Stereo-EEG and Intracranial EEG
Rapid seizure spread can obscure the beginning
Imagine a seizure genuinely starts in one small region.
Within seconds, it spreads into:
neighbouring cortex
the opposite hemisphere
and a wider network.
By the time the scalp EEG shows an obvious change, the original starting point may be difficult to identify.
The apparent scalp pattern may therefore describe where activity has spread, not necessarily the precise first neuronal event.
Interictal EEG and ictal EEG can also disagree
Interictal means between seizures.
Ictal means during the seizure itself.
Someone may have interictal epileptiform discharges predominantly from one area but actual seizures recorded from another or from a wider network.
Interictal abnormalities identify tissue capable of generating epileptiform activity.
They do not necessarily define the complete seizure-onset or epileptogenic zone.
This distinction is one reason prolonged video EEG is so important during surgical assessment.
Even interictal abnormalities can occur on both sides
A person with predominantly unilateral focal epilepsy may sometimes show interictal abnormalities involving:
both temporal lobes
several cortical regions
or a broader network.
The clinical meaning depends on:
how frequent each pattern is
the actual seizure recordings
MRI
symptoms
and the rest of the evaluation.
A bilateral interictal EEG finding therefore does not, by itself, settle whether focal surgery is impossible.
Why can PET show a larger abnormal area than EEG?
PET commonly examines brain metabolism between seizures.
The metabolic abnormality may extend beyond the precise tissue where seizures originate.
Long-standing epilepsy can affect connected networks.
Therefore PET hypometabolism may include:
the seizure-generating area
tissue involved in seizure propagation
and functionally connected regions.
PET is therefore valuable for localisation but should not be interpreted as:
“everything abnormal on PET must be removed.”
UCLH specifically uses PET to help identify possible seizure-related regions, particularly when MRI is normal, while combining it with the rest of the presurgical evidence.
PET can be particularly useful when the other results do not agree
Modern reviews describe FDG-PET as particularly useful when:
MRI is negative
or non-invasive investigations are discrepant.
PET can provide another independent piece of evidence about the likely seizure network.
It does not automatically overrule EEG or MRI.
Instead, it contributes another layer of evidence.
Why can SPECT appear different again?
Ictal SPECT examines blood-flow changes associated with a seizure.
But timing matters enormously.
The radioactive tracer must be injected rapidly.
If injection happens after the seizure has already spread, increased blood flow may reflect:
the original seizure-onset area
early propagation
or a mixture of both.
SPECT therefore needs careful interpretation alongside video EEG and seizure timing.
UCLH describes ictal SPECT as being compared with an interictal scan to identify seizure-associated changes in cerebral blood flow.
A large SPECT abnormality does not necessarily mean the seizure began throughout that whole area
Seizures spread.
Blood-flow changes can therefore extend through an active seizure network.
The specialist interpretation asks:
Which part of the pattern most plausibly reflects early seizure activity?
rather than simply treating every area of increased perfusion as the seizure origin.
Why can MEG disagree with EEG?
EEG and MEG both investigate neuronal activity, but they measure different physical signals.
EEG records electrical potentials at the scalp.
MEG records magnetic fields generated by neuronal currents.
The skull and other tissues affect electrical and magnetic signals differently.
MEG may therefore detect or localise some interictal activity differently from conventional scalp EEG.
Modern evidence supports its use as an additional modality particularly in MRI-negative focal epilepsy, but it still needs to be interpreted alongside other investigations.
MEG usually records activity between seizures
This is another important point.
Most clinical MEG localisation is based on interictal epileptiform discharges.
It therefore does not automatically identify exactly where an individual's clinical seizure begins.
It can provide valuable evidence about the epileptic network and can help guide later intracranial electrode placement.
Why can neuropsychology point somewhere different?
Neuropsychological assessment evaluates functions such as:
memory
language
attention
processing speed
executive function
and visual-spatial skills.
A pattern of difficulty may contribute evidence about which brain networks are affected.
But cognition is also influenced by:
seizure frequency
medication
sleep
mood
development
education
previous brain injury
and widespread network disruption.
Neuropsychology therefore contributes to lateralisation and localisation but should not be treated as a precise electrical seizure detector.
Current ILAE guidance describes neuropsychology as a collaborative contributor to seizure characterisation, lateralisation and localisation, while also establishing cognitive baselines and estimating surgical cognitive risk.
Poor memory does not automatically identify the seizure side
For example, memory difficulty might relate to:
temporal-lobe dysfunction
bilateral network effects
antiseizure medication
sleep disruption
frequent interictal activity
depression
anxiety
or several factors simultaneously.
A single low memory score cannot therefore prove:
“seizures come from this hippocampus.”
The pattern must be interpreted in clinical context.
Why can seizure symptoms appear to point somewhere different?
The person's symptoms are called seizure semiology.
They can provide powerful localisation clues.
But the first symptom the person notices is not always the first thing happening electrically in the brain.
A seizure may begin in a region producing no conscious symptom.
It may then spread to another area and produce:
déjà vu
fear
movement
sensory disturbance
language problems
or another noticeable experience.
The experienced symptom can therefore reflect early propagation rather than the exact electrical onset.
Memory can also hide the beginning
Some people do not remember the earliest part of a seizure.
For example, a seizure may begin with:
a subtle focal symptom
then impair consciousness
then progress to a convulsion.
The person may remember only:
“I suddenly woke up afterwards.”
Witness descriptions, video and EEG may reveal an earlier stage the person cannot report.
Why can a tonic-clonic seizure appear generalised even when it began focally?
Once a focal seizure spreads into both hemispheres, the later convulsive phase may appear completely bilateral.
If nobody saw the beginning, it may initially be classified as:
bilateral tonic-clonic seizure of unknown focal/generalised origin.
Later video EEG might reveal:
focal seizure → bilateral tonic-clonic seizure.
The apparent disagreement was therefore caused by missing the earliest part of the event, not because the seizure changed type.
What if MRI shows one side but seizures seem to come from the other?
This is an important discordance.
Possible explanations include:
the MRI lesion is incidental
scalp EEG localisation is misleading
seizure activity spreads rapidly
there are independent seizure networks
both hemispheres are involved
or the lesion and epileptogenic network are more complex than they initially appear.
A specialist team should not simply assume either the MRI or EEG is correct.
Further investigation may be needed.
What if MRI shows several abnormalities?
Multiple abnormalities create another problem.
Which one, if any, is epileptogenic?
Doctors may compare:
seizure symptoms
EEG localisation
PET
SPECT
MEG
and eventually intracranial EEG
to determine which abnormality best fits the seizure network.
The presence of more than one lesion can therefore increase rather than reduce the need for careful multimodal assessment.
What if MRI is normal but every other test points to one area?
This can happen.
A person may have:
highly stereotyped focal seizures
consistent scalp EEG localisation
concordant PET or MEG
and compatible neuropsychological findings
despite an apparently normal MRI.
Current NICE guidance specifically recommends that people with drug-resistant epilepsy be considered for epilepsy-surgery assessment even when MRI does not show an abnormality.
MRI-negative does not mean surgery is automatically impossible.
MRI-negative cases often require more investigation
A large 2025 systematic review found that people classified as MRI-negative were more likely to undergo SEEG and, as a group, had lower postoperative seizure-freedom rates than people with a visible relevant lesion.
The same review stressed that MRI-negative status itself is inconsistently defined and that modern specialist MRI review and post-processing can reveal abnormalities missed initially.
This is why MRI-negative epilepsy often requires a particularly careful multimodal evaluation.
What if EEG is normal but MRI shows a convincing lesion?
That can also happen.
Routine EEG has limited sensitivity.
NICE specifically states:
Do not use EEG to exclude a diagnosis of epilepsy.
A negative routine interictal EEG may simply mean that epileptiform activity was not captured during that particular recording.
A person with a structural lesion and convincing focal seizure history may therefore still have genuine epilepsy despite normal routine EEG.
What if EEG is abnormal but MRI is normal?
Again, this can be completely genuine.
The EEG measures functional electrical abnormality.
MRI looks for structural abnormality.
Epilepsy does not require a structural lesion visible on MRI.
The important question is whether the EEG abnormality:
fits the clinical events
supports the seizure classification
and behaves consistently across the wider evaluation.
What if the abnormal EEG does not fit the seizures?
That deserves careful review.
An interictal EEG abnormality may not necessarily explain every event the person experiences.
Someone might have:
epilepsy
plus syncope
functional seizures
migraine
parasomnias
or another disorder.
The presence of established epileptiform abnormalities should not automatically turn every later unusual event into an epileptic seizure.
Related Information Hub page:
Second Opinions and Re-evaluating an Epilepsy Diagnosis
Several different event types can create apparent contradictions
Suppose someone has:
Event A: brief stereotyped temporal-lobe seizures.
and:
Event B: a completely different prolonged shaking episode.
If Event A is captured on EEG while Event B is described only from memory, the overall history may initially appear inconsistent.
Recording each habitual event type separately can reveal that there is more than one process occurring.
This is one reason video telemetry sometimes needs to capture each distinct habitual event, not merely one seizure.
Epilepsy can also genuinely involve more than one network
Not every discordant result is caused by technical limitation.
Some people really do have:
multifocal epilepsy
independent seizures from both temporal lobes
widespread structural disease
or complex epileptic networks.
In such cases, tests pointing to several regions may accurately reflect the biology.
The challenge is determining whether there is still a dominant treatable network.
Bilateral does not automatically mean generalised epilepsy
This distinction is crucial.
A focal epilepsy can produce:
bilateral interictal abnormalities
rapid bilateral seizure spread
or independent focal abnormalities on both sides.
That is not the same as a primary generalised epilepsy network.
Classification therefore requires:
seizure semiology
EEG pattern
syndrome features
imaging
and the wider clinical context.
Related Information Hub page:
Focal, Generalised, Unknown and Unclassified Seizures — Understanding the 2025 ILAE Classification
Why test quality matters
Before interpreting discordance as biological, specialists should consider whether one investigation was technically limited.
Examples include:
movement during MRI
inadequate MRI protocol
insufficient EEG electrode coverage
muscle artefact during a seizure
no habitual event being captured
delayed SPECT injection
poor fMRI task performance
or limited neuropsychological validity.
A technically inadequate test should not be given the same weight as a high-quality result.
MRI interpretation expertise matters too
A subtle focal cortical dysplasia might be missed on an ordinary scan review and become apparent only when:
epilepsy specialists know where seizures are suspected to begin
the scan is reviewed by an experienced neuroradiologist
or specialist post-processing is performed.
NICE specifically recommends neuroradiology expertise in MRI interpretation and additional tertiary-centre review when seizures remain ongoing and diagnosis is unclear.
Knowing the EEG result can help MRI review
This may sound like bias, but in specialist epilepsy imaging it can be clinically useful.
If neuroradiologists know that several seizures consistently appear to begin in a particular region, they can scrutinise that region for subtle abnormalities.
The important safeguard is that the imaging still needs objective expert interpretation.
Clinical and imaging evidence should inform each other without forcing one result to match another.
Different tests also have different spatial resolution
No test sees the brain at exactly the same scale.
Scalp EEG measures summed electrical activity visible from outside the head.
MRI shows anatomical structure in great detail.
PET and SPECT display metabolic or perfusion patterns that may cover comparatively broad areas.
SEEG provides very detailed electrical information from selected implanted locations, but does not record the whole brain.
Therefore apparent differences in size or boundary are expected.
Different tests have different temporal resolution too
EEG can follow electrical activity over milliseconds.
PET represents metabolism over a much longer period.
SPECT effectively reflects perfusion around the time the tracer is taken up.
MRI generally represents anatomy rather than the moment-to-moment seizure.
Neuropsychology reflects cognitive performance over an assessment session.
These investigations are not taking photographs of the exact same biological moment.
SEEG can resolve some disagreements — but not all
When non-invasive investigations leave a small number of plausible focal hypotheses, SEEG can record directly from selected regions.
It may establish that:
one region consistently becomes ictal first
apparent bilateral activity actually spreads from one side
several independent onset zones exist
or the original hypothesis was incorrect.
But SEEG samples only the areas where electrodes were placed.
A region not sampled cannot be ruled out with absolute certainty.
Related Information Hub page:
Stereo-EEG and Intracranial EEG
SEEG itself can be discordant with earlier tests
Suppose scalp EEG appeared left temporal but SEEG records seizures from the insula.
That does not necessarily mean the scalp EEG was “bad”.
The insular seizure may have spread rapidly to temporal cortex, creating a temporal scalp pattern.
The invasive recording provided higher-resolution information closer to the source.
The earlier scalp result may still have accurately recorded later propagation.
What does the MDT do when tests disagree?
The multidisciplinary team does not simply count votes.
It would be unsafe to say:
MRI says left, EEG says right, PET says left — left wins 2–1.
Instead, specialists consider:
what each test actually measured
technical quality
whether it examined seizure onset or interictal activity
how specific the finding is
whether the result could reflect propagation
whether the abnormality may be incidental
and how well it fits the clinical seizure pattern.
Different evidence has different weight in different cases.
The clinical question determines which result matters most
For example:
If the question is:
“Is there a subtle structural lesion?”
specialist MRI may carry particular importance.
If the question is:
“Where does the habitual seizure begin electrically?”
ictal EEG or SEEG may be more directly informative.
If the question is:
“Where is language organised?”
functional mapping is more relevant.
If the question is:
“What is the likely cognitive risk of surgery?”
neuropsychology becomes particularly important.
There is no universal hierarchy in which one test always defeats all others.
Sometimes the correct response to disagreement is another test
Further investigation might include:
specialist MRI review
repeat epilepsy-protocol MRI
additional video telemetry
PET
ictal SPECT
MEG
functional MRI
neuropsychological reassessment
or intracranial EEG.
UCLH describes PET and SPECT as additional investigations used particularly when MRI and EEG have not adequately pinpointed seizure origin, and intracranial EEG when still greater precision is needed.
Additional testing should answer a question
More testing is not automatically better.
A new investigation should ideally answer something specific.
For example:
Is this MRI lesion actually part of the epileptic network?
Is the apparent bilateral scalp activity spreading from one side?
Which of two competing regions should SEEG sample?
Does functional imaging support the suspected MRI-negative focus?
Testing without a defined clinical question can produce additional findings without resolving the original uncertainty.
Sometimes repeating a test is more useful than adding a different test
If the original MRI was:
old
low-resolution
affected by movement
or not performed with an appropriate epilepsy protocol,
repeating MRI may provide more useful information than immediately ordering an entirely different investigation.
NICE specifically recommends considering additional MRI when:
the original scan was suboptimal
epilepsy develops new features
certain presumed non-structural epilepsies fail first-line treatment
or surgery is being considered.
Sometimes re-reading old tests changes the picture
The most useful new information may come from existing data.
After several seizures have been recorded, specialists may revisit an old MRI knowing exactly which region deserves particular scrutiny.
Or:
a video-EEG recording may be reviewed after a new seizure type has been recognised.
Reinterpretation can sometimes connect pieces of evidence that originally appeared unrelated.
What if PET, SPECT or MEG point towards an abnormality that MRI cannot see?
This can provide a target for closer investigation.
Functional imaging may help:
guide specialist MRI review
guide MRI post-processing
identify regions for intracranial sampling
or strengthen an MRI-negative surgical hypothesis.
A 2025 review of nuclear imaging in epilepsy specifically describes PET and SPECT as particularly relevant when epilepsy is non-lesional or when structural and electrophysiological evidence is discordant.
Functional imaging does not automatically turn MRI-negative epilepsy into MRI-positive epilepsy
A PET abnormality is not an MRI lesion.
If MRI remains structurally negative but PET identifies focal hypometabolism, the person still has MRI-negative epilepsy in the structural imaging sense.
What has improved is the multimodal localisation evidence.
The distinction matters.
Can surgery still happen when every test is not perfectly concordant?
Sometimes, yes.
Perfect agreement is not always achievable.
The team may still have enough high-quality evidence to support a treatment hypothesis despite some non-localising or less important conflicting findings.
Older and contemporary surgical literature both recognise that individual modalities may be non-localising or imperfect while the overall integrated evidence remains sufficient.
The decision depends on:
nature of the discordance
quality of evidence
surgical target
likely benefit
and risk.
But important unexplained discordance can stop or delay surgery
If major evidence suggests genuinely incompatible seizure origins, the team may decide that the risk of operating on the wrong network is too high.
Possible outcomes include:
further non-invasive investigation
SEEG
continuing medical treatment
neuromodulation
or deciding that resective surgery cannot currently be recommended.
This is not necessarily a failed assessment.
It may be the safest conclusion available from the evidence.
Why should surgery sometimes be delayed rather than “trying” the most likely area?
Because brain surgery is irreversible.
Removing tissue based on insufficient localisation can:
fail to control seizures
produce neurological deficits
and complicate future treatment.
Presurgical evaluation therefore aims for sufficient confidence before treatment, not merely a best guess.
Concordance also matters for predicting surgical success
Multimodal concordance generally strengthens confidence that treatment is targeting the correct epileptogenic network.
A recent systematic review of mesial temporal epilepsy with hippocampal sclerosis found that concordant MRI, EEG and PET evidence was among the strongest predictors of favourable seizure outcomes within that specific patient group.
However, this should not be converted into a universal rule that discordant findings mean surgery will fail.
Different epilepsy types have different evidence and outcomes.
MRI-negative surgery can still succeed
The absence of a visible lesion generally makes localisation more difficult, but it does not make successful treatment impossible.
A 2025 systematic review found favourable postoperative outcomes in a substantial proportion of MRI-negative patients, although seizure-freedom rates were lower overall than in MRI-positive groups.
This is precisely why modern multimodal investigation remains important.
A test can be correct but answer a different question
This is perhaps the simplest way to understand apparent disagreement.
MRI may correctly show:
where abnormal tissue exists.
EEG may correctly show:
where electrical activity is visible at the scalp.
PET may correctly show:
where metabolism is altered between seizures.
Neuropsychology may correctly show:
which cognitive networks are functioning poorly.
SEEG may correctly show:
which implanted contact recorded the earliest detectable ictal change.
Those statements are related.
They are not identical.
“Which test is the most accurate?” is often the wrong question
A better question is:
“Accurate for what?”
No single epilepsy investigation is best at answering every question.
Doctors therefore choose investigations according to whether they need to understand:
diagnosis
seizure classification
structural cause
electrical onset
seizure spread
metabolic dysfunction
functional anatomy
cognitive risk
or suitability for surgery.
What if tests disagree outside surgical assessment?
Discordance matters even when surgery is not being considered.
For example:
a person diagnosed with generalised epilepsy may later develop a consistently focal seizure pattern;
or:
EEG may suggest focal abnormalities while the clinical history sounds very different;
or:
a structural lesion may be discovered that does not fit the existing diagnosis.
This may prompt reconsideration of:
seizure classification
epilepsy syndrome
underlying cause
medication choice
or the diagnosis itself.
Treatment response can also provide new information
If seizures continue despite an appropriate medicine, NICE recommends reviewing the epilepsy diagnosis before simply continuing along the same treatment pathway.
Persistent treatment failure may suggest:
genuine drug-resistant epilepsy
incorrect classification
more than one seizure type
an alternative event type
or a more complicated epilepsy network.
Treatment response is not itself a diagnostic test, but it can reveal when the existing explanation needs another look.
What should a patient ask when results appear contradictory?
Useful questions include:
Which results agree with each other?
Which results are considered discordant?
Do the tests actually contradict each other, or are they measuring different parts of the same network?
Which finding do you consider most reliable, and why?
Could the scalp EEG be showing seizure spread rather than the true onset?
Could the MRI abnormality be incidental?
Has my MRI been reviewed by a specialist epilepsy neuroradiologist?
Would repeat MRI or post-processing be useful?
Does PET, SPECT or MEG support either hypothesis?
Do I need additional video EEG?
Would SEEG help resolve the disagreement?
Questions before surgery when tests do not fully agree
Useful questions include:
How confident are you about where my seizures begin?
Which investigations provide the strongest evidence?
Which finding worries the MDT most?
What are the possible explanations for the discordance?
What would SEEG need to prove?
What happens if SEEG does not confirm the current hypothesis?
Would surgery still be considered if one test remains discordant?
How does this uncertainty affect my expected chance of seizure improvement?
Could the discordance increase the neurological risks of surgery?
Are there non-resective treatment options if localisation remains uncertain?
The most important message
Epilepsy tests sometimes disagree because they are not all measuring the same thing.
MRI examines structure.
EEG records electrical activity.
PET examines metabolism.
SPECT examines blood flow.
MEG records magnetic fields associated with neuronal activity.
Neuropsychology examines cognitive performance.
Functional MRI maps task-related brain activity.
SEEG records electrical activity directly from selected implanted regions.
Each technique provides a different view of the epilepsy network.
A result may also be affected by:
seizure spread
timing
technical quality
limited sampling
deep seizure origin
individual brain organisation
or the fact that epilepsy genuinely involves several regions.
This means an apparent disagreement does not automatically prove that one test is wrong.
The important question is whether specialists can build a coherent explanation that accounts for all of the evidence.
When several independent investigations support the same explanation, this is called concordance.
When important findings point in different directions, this is discordance.
Discordance may lead to:
specialist re-review
repeat testing
PET
SPECT
MEG
additional video EEG
functional assessment
or SEEG.
Sometimes the eventual conclusion is that surgery can proceed.
Sometimes further investigation resolves the disagreement.
And sometimes uncertainty remains too great for a safe resection.
That last outcome does not mean the investigation failed.
In epilepsy care—particularly before irreversible brain surgery—recognising when the evidence does not yet provide a sufficiently reliable answer can be just as important as identifying a clear surgical target.
Sources and further reading
NICE — Epilepsies in children, young people and adults (NG217): Diagnosis and assessment.
Current NICE guidance states that EEG must not be used to exclude epilepsy, MRI should use epilepsy protocols and specialist neuroradiology expertise, and additional MRI review or repeat MRI should be considered in selected people when seizures remain ongoing or surgery is being considered. Last updated January 2025.
NICE — Rationale and impact for NG217.
Explains the limitations of routine interictal EEG, the role of specialist MRI interpretation and the recommendation that people with drug-resistant epilepsy be considered for surgical assessment even when MRI has not identified an abnormality.
University College London Hospitals — Epilepsy investigations and presurgical assessment.
Describes the complementary roles of MRI, fMRI, PET, SPECT, neuropsychology and intracranial EEG, including the use of PET in MRI-negative cases and SPECT when MRI and EEG have not adequately localised seizures.
International League Against Epilepsy — Epilepsy Surgery Commission.
Current ILAE work includes presurgical evaluation, intracranial EEG and the analysis of epileptogenic brain networks. The commission's 2026 reports include systematic work on intracranial EEG connectivity for localisation of epileptogenic networks.
MRI-negative epilepsy: systematic review and meta-analysis, 2025.
Reviewed more than 10,000 MRI-negative cases and highlighted variability in how MRI-negative epilepsy is defined, the value of MRI post-processing and the greater use of SEEG in MRI-negative surgical assessment.
Preoperative predictors of seizure outcomes after surgery for mesial temporal sclerosis: systematic review, 2026.
Found that multimodal concordance involving MRI-visible hippocampal sclerosis, unilateral EEG seizure onset and concordant PET findings was strongly associated with favourable postoperative outcomes in the population studied.
Nuclear Medicine Imaging in Epilepsy, 2025.
Reviews the complementary roles of PET and SPECT, particularly in MRI-negative epilepsy, multiple-lesion cases and situations where electrophysiological and structural findings are discordant.
ILAE — Neuropsychological assessment in epilepsy surgery, 2026.
Current ILAE guidance describes neuropsychology as contributing collaboratively to seizure characterisation, localisation and lateralisation while establishing cognitive baselines and predicting postoperative cognitive risk.
Information reviewed: September 2026.
This page provides general educational information. Apparently conflicting epilepsy investigations require specialist interpretation because the significance of discordance depends on the individual seizure pattern, quality of each investigation and the clinical question being asked.
