Advanced Imaging: PET, SPECT & MEG

When structural MRI alone is insufficient to identify the epileptogenic zone — as occurs in up to 20–30% of patients with focal epilepsy — advanced functional and metabolic imaging modalities become essential components of the presurgical evaluation. Fludeoxyglucose positron emission tomography (FDG-PET), ictal single-photon emission computed tomography (SPECT), magnetoencephalography (MEG), and functional MRI (fMRI) each provide unique and complementary information about epileptic networks, metabolic activity, and eloquent cortex. These modalities are most powerful when used in combination with structural MRI and electrophysiologic data to build a concordant hypothesis about the epileptogenic zone. Understanding the principles, strengths, and limitations of each technique allows the epileptologist to select the appropriate imaging studies for individual patients and interpret their results in clinical context.

Bottom Line

FDG-PET identifies regions of interictal hypometabolism corresponding to the seizure network; particularly valuable in MRI-negative epilepsy, where concordant PET hypometabolism predicts favorable surgical outcome
Ictal SPECT detects seizure-related hyperperfusion but requires injection during the seizure; SISCOM (subtraction ictal SPECT coregistered to MRI) enhances localization by quantitatively comparing ictal to interictal scans
MEG / MSI provides millisecond temporal resolution for source localization of interictal discharges; complementary to EEG for detecting tangentially oriented sources and can localize epileptiform activity in MRI-negative cases
fMRI noninvasively maps eloquent cortex (language, motor, memory) for presurgical planning; the 2017 AAN guideline supports fMRI for language lateralization and, as an option (Level B), for predicting verbal memory decline — but found evidence insufficient to completely replace the Wada test except possibly for language lateralization
Concordance across modalities is the strongest predictor of good surgical outcome; discordant results warrant intracranial EEG or reconsideration of the surgical hypothesis
Emerging techniques — 7T MRI, PET-MRI hybrid scanners, EEG-fMRI, and diffusion tensor imaging — are pushing the boundaries of noninvasive epilepsy evaluation

FDG-PET

Fludeoxyglucose positron emission tomography (FDG-PET) has been used in the surgical evaluation of drug-resistant focal epilepsy for over 40 years and remains one of the most widely available and clinically validated advanced imaging modalities in epileptology.

Principles and Technique

Radiotracer: 18F-labeled 2-fluoro-2-deoxy-D-glucose (FDG) is the most common ligand due to its wide availability at PET imaging centers
Mechanism: FDG is taken up by metabolically active cells and phosphorylated but cannot proceed through glycolysis, trapping the radiotracer proportionally to the rate of glucose metabolism
Timing: Performed interictally (no seizure for at least 24 hours before injection); the patient rests quietly in a dimly lit room during the 30–60 minute uptake period to minimize non-epileptic metabolic variation
Imaging: PET scan acquired 30–60 minutes after injection; images coregistered to the patient’s structural MRI for anatomic correlation

Interpretation

Finding	Significance	Clinical Implication
Focal hypometabolism	Region of the seizure network with decreased glucose metabolism interictally	Localizes to the hemisphere and lobe of seizure onset; the region of hypometabolism often extends beyond the actual seizure onset zone
Concordant with MRI lesion	Hypometabolism corresponding to the structural lesion	Strengthens the surgical hypothesis; supports proceeding with resection without intracranial EEG in select cases
Positive in MRI-negative epilepsy	Hypometabolism without a visible structural lesion	Particularly valuable; surgical outcomes in PET-positive, MRI-negative temporal lobe epilepsy may be similar to those with mesial temporal sclerosis on MRI
Bilateral or widespread hypometabolism	Extensive seizure network involvement; possible bilateral pathology	May suggest more complex epilepsy; warrants careful electrophysiologic evaluation before surgery

FDG-PET in MRI-Negative Epilepsy

FDG-PET has been demonstrated to be valuable in both lesional and nonlesional epilepsy, but it is especially helpful in MRI-negative cases
In a large study, Moon and colleagues found that concordant FDG-PET was a significant independent prognostic indicator for good surgical outcome in nonlesional focal epilepsy
Multiple studies have shown that patients with PET-positive, MRI-negative mesial temporal lobe epilepsy have similar surgical outcomes compared with those who have mesial temporal sclerosis on MRI
FDG-PET hypometabolism can guide placement of intracranial electrodes when the MRI is non-localizing
Limitation: The region of hypometabolism is typically larger than the actual epileptogenic zone, reflecting network effects rather than the precise seizure focus

Ictal SPECT

Single-photon emission computed tomography (SPECT) is used to study cerebral perfusion changes during seizures. The technique exploits the coupling between neuronal activity and regional cerebral blood flow: the cortex involved in seizure onset becomes hyperperfused relative to baseline.

Technique

Radiotracer: Technetium-99m (Tc-99m)-labeled hexamethylpropyleneamine oxime (HMPAO) or ethyl cysteinate dimer (ECD); these lipophilic tracers cross the blood-brain barrier and are trapped in proportion to regional cerebral blood flow
Injection timing: The radiotracer must be injected during the seizure — ideally within the first 20–30 seconds of electrographic onset. This is the most critical aspect of the technique; any delay may result in imaging of seizure propagation rather than the onset zone
Setting: Performed in a video-EEG monitoring unit with specially trained staff approved to handle radioactive material; the radiotracer is kept at the bedside (autoinjector or manual injection by trained technician); IV access must be functioning
Image acquisition: The SPECT scan can be obtained hours after injection (the tracer is “fixed” in the brain at the moment of injection) — the patient does not need to be in the scanner during the seizure

Interictal SPECT

A baseline interictal SPECT scan is also performed (during a seizure-free period) for qualitative comparison with the ictal scan
The interictal scan typically shows no focal perfusion abnormalities or may show subtle hypoperfusion in the epileptogenic region

SISCOM (Subtraction Ictal SPECT Coregistered to MRI)

SISCOM is a quantitative post-processing technique that significantly enhances the clinical utility of SPECT imaging.

Method: The interictal SPECT image is digitally subtracted from the ictal SPECT image, and the resulting difference map (showing areas of increased perfusion during the seizure) is coregistered to the patient’s structural MRI
Advantage: Provides precise anatomic localization of the zone of maximal ictal hyperperfusion; more accurate than visual comparison of ictal and interictal SPECT alone
Utility: SISCOM has greater utility in temporal lobe epilepsy compared with extratemporal lobe epilepsy, though it can be informative in both
Concordance: When the SISCOM localization is concordant with other data (EEG, MRI, PET), it strengthens the surgical hypothesis and may reduce the need for intracranial EEG

Limitations and Pitfalls of Ictal SPECT

Injection delay: The most common source of false localization; if injection occurs more than 20–30 seconds after seizure onset, the SPECT may image seizure propagation rather than onset, potentially pointing to the wrong region
Rapid EEG spread: Seizures that spread quickly (e.g., from temporal to frontal lobe within seconds) may produce SPECT images showing hyperperfusion in the propagation zone rather than the onset zone
Short seizures: Brief seizures (<10 seconds) may end before sufficient radiotracer is trapped, producing inadequate or misleading images
Postictal changes: If the injection occurs after the seizure has ended (even by seconds), postictal hypoperfusion may be captured instead of ictal hyperperfusion, leading to false lateralization
Practical challenges: Maintaining a functioning IV and having trained staff available 24/7 to inject during an unpredictable event is resource-intensive

Magnetoencephalography (MEG)

Magnetoencephalography directly measures the magnetic fields generated by neuronal electrical activity. Because magnetic fields are less distorted by the skull and scalp than electrical potentials, MEG offers superior spatial resolution for source localization compared with scalp EEG.

Principles

Signal source: MEG detects the magnetic fields produced by intracellular postsynaptic currents flowing in the apical dendrites of pyramidal neurons; these are the same currents that generate the EEG signal
Sensor technology: Superconducting quantum interference devices (SQUIDs) housed in a magnetically shielded room detect extremely weak biomagnetic signals (femtotesla range)
Temporal resolution: Millisecond-level, comparable to EEG; far superior to the temporal resolution of PET, SPECT, or fMRI
Spatial resolution: Superior to EEG for source localization, particularly for tangentially oriented dipoles (e.g., those arising from the walls of sulci)

Magnetic Source Imaging (MSI)

Definition: MSI refers to the integration of MEG dipole source analysis with the patient’s structural MRI, producing a map of epileptiform source locations on the brain surface
Applications:
- Localization of interictal epileptiform discharges, especially when EEG localization is ambiguous or multifocal
- Detection of epileptiform sources in MRI-negative epilepsy
- Guidance for intracranial electrode placement
- Complementary to EEG: MEG is most sensitive to tangentially oriented sources (sulcal), while EEG is more sensitive to radially oriented sources (gyral crests)

Feature	EEG	MEG
Signal source	Extracellular ionic currents (volume conduction)	Intracellular postsynaptic currents (magnetic fields)
Sensitivity to dipole orientation	Radial > tangential	Tangential > radial
Effect of skull/scalp	Significant attenuation and smearing	Minimal (magnetic fields pass through tissue with little distortion)
Spatial resolution (source localization)	~10–20 mm (scalp EEG)	~5–10 mm (with appropriate dipole modeling)
Temporal resolution	Millisecond	Millisecond
Recording environment	Standard clinical room	Magnetically shielded room (expensive infrastructure)
Availability	Ubiquitous	Limited to specialized epilepsy centers
Deep sources	Detectable (but attenuated)	Less sensitive to deep (e.g., mesial temporal) sources

Functional MRI (fMRI)

Functional MRI is performed to noninvasively map eloquent cortical regions onto the brain surface, typically for presurgical planning. It has become a standard component of the epilepsy surgery workup for assessing the risk of postoperative neurologic deficits.

Principles

BOLD signal: fMRI measures the blood oxygen level–dependent (BOLD) effect on T2*-weighted acquisitions. Increased neuronal activity leads to increased local cerebral blood flow, which paradoxically increases the ratio of oxygenated to deoxygenated hemoglobin (the metabolic demand is overcompensated), altering the local magnetic field and producing the BOLD signal
Task-based paradigm: The patient performs specific cognitive, motor, or language tasks during image acquisition. Common tasks include verb generation (Broca area), reading comprehension (Wernicke area), finger tapping (motor cortex), and word encoding (hippocampal activation)
Resting-state fMRI: An emerging approach that analyzes intrinsic brain network connectivity without task performance; potentially useful for patients who cannot cooperate with task paradigms

Clinical Applications

Application	Clinical Question	Evidence Level
Language lateralization	Which hemisphere is dominant for language? Is surgery likely to cause aphasia?	Best evidence; AAN guideline supports use for language lateralization, especially in temporal lobe epilepsy; may replace the Wada test in many centers
Verbal memory prediction	Will temporal lobectomy cause clinically significant memory decline?	AAN guideline (Level B): fMRI should be considered as an option to predict verbal memory decline after left MTL surgery; leftward activation asymmetry probably predicts postoperative verbal memory outcome
Motor cortex mapping	How close is the epileptogenic zone to the primary motor cortex?	Useful for perirolandic epilepsy surgery planning; complements direct cortical stimulation
Visual cortex mapping	Will occipital resection cause visual field deficits?	Can map primary and association visual cortex; combined with diffusion tractography for optic radiation assessment

fMRI vs. Wada Test (Intracarotid Amobarbital Procedure)

The Wada test was historically the gold standard for language and memory lateralization before epilepsy surgery
fMRI has increasingly replaced the Wada test at many epilepsy centers due to its noninvasive nature, repeatability, and ability to provide spatial localization (not just lateralization)
The AAN practice guideline (2017) concluded that fMRI is helpful for language lateralization, particularly in temporal lobe epilepsy, and may be used to predict verbal memory decline
Wada test is still used when fMRI is inconclusive, when patients cannot cooperate with fMRI tasks, or when bilateral language representation is suspected

Diffusion Tensor Imaging (DTI) and White Matter Tractography

Diffusion tensor imaging is an MRI technique that measures the directional diffusion of water molecules along white matter tracts, allowing three-dimensional reconstruction of major fiber pathways.

Principles: Water diffuses preferentially along the axis of myelinated axons (anisotropic diffusion); DTI quantifies this directionality using fractional anisotropy (FA) and mean diffusivity (MD) metrics
Tractography: Computational algorithms trace fiber pathways between regions of interest, generating 3D reconstructions of tracts such as the optic radiations, corticospinal tract, and arcuate fasciculus
Clinical applications in epilepsy surgery:
- Optic radiation mapping: Predicting visual field deficits before temporal lobectomy (Meyer’s loop); critical for occipital lobe surgery planning
- Corticospinal tract mapping: Assessing risk of motor deficits in perirolandic or insular epilepsy surgery
- Language tract mapping: Arcuate fasciculus and other language-associated tracts; complements fMRI language mapping
- Identifying white matter disruption: Reduced FA in white matter adjacent to FCD or cortical tubers can help identify the epileptogenic lesion

When to Order Each Modality

Modality	Best Indication	Key Advantage	Key Limitation
FDG-PET	MRI-negative focal epilepsy; concordance assessment for presurgical evaluation	Widely available; interictal study (no seizure needed); concordant PET predicts good surgical outcome	Hypometabolism extends beyond the seizure onset zone; limited spatial resolution (~6–8 mm)
Ictal SPECT / SISCOM	Extratemporal epilepsy localization; MRI-negative or discordant cases	Captures seizure-related perfusion changes; SISCOM provides anatomic overlay	Requires injection during seizure; injection delay leads to false localization; resource-intensive
MEG / MSI	MRI-negative epilepsy; ambiguous EEG localization; sulcal/tangential sources	Superior spatial resolution; complements EEG (detects tangential dipoles missed by EEG)	Limited availability; expensive; less sensitive to deep sources; interictal only
fMRI	Presurgical language lateralization; memory prediction; eloquent cortex mapping	Noninvasive; repeatable; spatial localization (not just lateralization); may replace Wada test	Requires patient cooperation; susceptibility to motion artifact; indirect measure of neuronal activity
DTI / tractography	Presurgical planning near eloquent tracts (optic radiations, CST, arcuate fasciculus)	3D visualization of white matter pathways at risk during surgery	Cannot determine functional significance of a tract; fiber crossing limitations; not standardized across centers

Emerging Techniques

7T MRI

Ultra-high-field 7T MRI provides approximately double the signal-to-noise ratio of 3T, enabling visualization of hippocampal subfields, cortical layers, and microstructural abnormalities not visible on conventional imaging
Preliminary studies in epilepsy have demonstrated improved detection of subtle FCD, hippocampal sclerosis, and polymicrogyria compared with 3T imaging
Limitations: Limited availability; increased susceptibility artifacts; specific absorption rate constraints; not yet approved for clinical use in all regions

PET-MRI Hybrid Scanners

Combined PET-MRI scanners acquire simultaneous PET and MRI data in a single session, providing perfectly coregistered structural and metabolic imaging
Potential advantages: reduced imaging time; improved coregistration accuracy; reduced radiation exposure (compared with PET-CT); simultaneous multimodal data acquisition
Early studies suggest comparable or superior lesion detection to sequential PET and MRI in epilepsy evaluation

EEG-fMRI

Simultaneous recording of EEG inside the MRI scanner allows mapping of BOLD signal changes time-locked to interictal epileptiform discharges
Identifies the hemodynamic correlate of specific epileptiform patterns, potentially revealing the location of the epileptogenic zone
Technically demanding (EEG artifact from MRI gradient switching must be carefully removed); primarily used in research settings but emerging in clinical practice

Post-Processing and Machine Learning

Automated morphometric analysis (voxel-based morphometry, cortical thickness analysis, FLAIR intensity normalization) can detect subtle FCD and other lesions missed on visual review
Machine learning classifiers trained on large datasets of lesional and non-lesional MRIs are showing promise in automated FCD detection, with some studies reporting sensitivities of 70–85% for lesions missed by expert neuroradiologists
These tools are expected to become standard components of epilepsy imaging pipelines in the coming years

Multimodal Integration

The most effective presurgical evaluations integrate data from multiple imaging modalities with electrophysiologic findings to construct a coherent hypothesis about the epileptogenic zone.

Concordance is key: When MRI, EEG, PET, SPECT, and MEG all point to the same region, surgical outcomes are excellent and intracranial EEG may not be necessary
Discordance requires further evaluation: When noninvasive data are discordant, intracranial EEG (stereo-EEG or subdural grids) is typically required to resolve the conflicting information before surgery
No single modality is sufficient alone: Each technique has blind spots (EEG misses deep sources; MRI misses subtle FCD; PET overestimates the epileptogenic zone; SPECT is injection-timing dependent). The power lies in their combination

References

Bernasconi A, Cendes F, Theodore WH, et al. Recommendations for the use of structural magnetic resonance imaging in the care of patients with epilepsy. Epilepsia. 2019;60(6):1054–1068.
Moon HJ, Kim DW, Chung CK, et al. Change of patient selection strategy and improved surgical outcome in MRI-negative neocortical epilepsy. J Epilepsy Res. 2016;6(2):66–74.
Carne RP, O’Brien TJ, Kilpatrick CJ, et al. MRI-negative PET-positive temporal lobe epilepsy: a distinct surgically remediable syndrome. Brain. 2004;127(Pt 10):2276–2285.
LoPinto-Khoury C, Sperling MR, Skidmore C, et al. Surgical outcome in PET-positive, MRI-negative patients with temporal lobe epilepsy. Epilepsia. 2012;53(2):342–348.
Yoganathan K, Malek N, Torzillo E, Paranathala M, Greene J. Neurological update: structural and functional imaging in epilepsy surgery. J Neurol. 2023;270(5):2798–2808.
Szaflarski JP, Gloss D, Binder JR, et al. Practice guideline summary: use of fMRI in the presurgical evaluation of patients with epilepsy. Neurology. 2017;88(4):395–402.
Semah F, Picot MC, Adam C, et al. Is the underlying cause of epilepsy a major prognostic factor for recurrence? Neurology. 1998;51(5):1256–1262.
Skidmore CT. Adult focal epilepsies. Continuum (Minneap Minn). 2016;22(1, Epilepsy):94–115.
Lapalme-Remis S. Neuroimaging of epilepsy. Continuum (Minneap Minn). 2022;28(2):306–338.
Hadady L, Sperling MR, Alcala-Zermeno JL, et al. Prediction tools and risk stratification in epilepsy surgery. Epilepsia. 2024;65(2):414–421.
Nune G, Arcot Desai S, Razavi B, et al. Treatment of drug-resistant epilepsy in patients with periventricular nodular heterotopia using RNS System. Clin Neurophysiol. 2019;130(8):1196–1207.
Knake S, Triantafyllou C, Wald LL, et al. 3T phased array MRI improves the presurgical evaluation in focal epilepsies. Neurology. 2005;65(7):1026–1031.
Zhu H, Scott J, Hurley A, et al. 1.5 versus 3 tesla structural MRI in patients with focal epilepsy. Epileptic Disord. 2022;24(2):274–286.
Winston GP, Micallef C, Kendell BE, et al. The value of repeat neuroimaging for epilepsy at a tertiary referral centre. Epilepsy Res. 2013;105(3):349–355.
Blümcke I, Thom M, Aronica E, et al. The clinicopathologic spectrum of focal cortical dysplasias. Epilepsia. 2011;52(1):158–174.