Improving MRI-Negative Epilepsy Localization: Synergy of Dual-Probe PET/MR (¹⁸F-FDG/¹¹C-FMZ)

Refractory epilepsy, characterized by resistance to two or more appropriately selected and dosed antiepileptic drugs^1,2,3, often presents as focal epilepsy accompanied by neurocognitive impairment and comorbid mental disorders^4,5,6. When pharmacological treatments fail, surgical resection of the epileptogenic focus becomes a critical intervention, making precise preoperative localization a key prerequisite for successful outcomes.

Conventional magnetic resonance imaging (MRI) fails to identify the epileptogenic focus in approximately 30-40% of patients with refractory epilepsy⁷. Standard protocols, often employing a 1.5 T scanner and 5 mm slice thickness, can miss subtle lesions like hippocampal sclerosis or focal cortical dysplasia⁸. Moreover, structural MRI cannot capture the dynamic metabolic changes that occur between interictal and ictal states⁹, underscoring the limited sensitivity of conventional imaging. Although electroencephalography (EEG) can provide electrophysiological localization information, its spatial resolution is limited, and it is easily affected by scalp tissue attenuation effects¹⁰. These limitations make it difficult to achieve accurate three-dimensional localization of the epileptogenic focus. Moreover, invasive procedures such as intracranial EEG monitoring are often required^11,12. These factors collectively lead to prolonged preoperative assessment cycles for MRI-negative epilepsy patients, increasing the risks associated with surgical decision-making and directly affecting clinical prognosis.

The clinical application of integrated positron emission tomography/magnetic resonance imaging (PET/MR) technology offers a promising solution to this challenge. This technology can simultaneously acquire and functional and metabolic information, as well as high-resolution anatomical structure data, enabling precise spatial and temporal fusion of multimodal images. While PET/MR inherently improves workflow efficiency by integrating multiple exams into a single scan, there are still limitations in the diagnostic specificity of a single radiotracer¹³. ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) PET often reveals extensive hypometabolic regions, which may be beyond the scope of the true epileptogenic focus, leading to blurring of the surgical border¹⁴. Moreover, single functional imaging technologies, such as 18F-FDG PET, still carry the risk of false negatives¹⁵. This limitation is mainly due to their insufficient spatial resolution, which restricts their ability to detect subtle lesions, such as microcortical dysplasia and focal cortical structural dysplasia¹⁶.

Studies indicate that combining FDG and ¹¹C-flumazenil (¹¹C-FMZ) significantly reduces the false-positive rate in localization. FMZ precisely delineates the core region of receptor abnormalities, while FDG may reflect a broader network of functional suppression^17,18. This study aims to establish a dual-probe PET/MR imaging protocol using ¹⁸F-FDG and ¹¹C-FMZ. In epileptic patients during the interictal period, epileptogenic foci show low metabolism because synaptic activity is suppressed¹⁹. 18F-FDG PET imaging reflects these regions of abnormal glucose metabolism^20,21,22,23. Meanwhile, 11C-FMZ PET imaging targets the distribution of GABA_A receptors, enabling specific identification of the epileptogenic focus, which has abnormal receptor density²⁴. This dual-probe PET/MR technique simultaneously acquires functional and anatomical information, thereby reducing spatiotemporal-registration errors. Combining metabolic (FDG) and receptor (FMZ) dual-parameter analysis enhances diagnostic specificity. It not only improves the accuracy of epileptogenic focus localization but also allows assessment of the functional connectivity features of the epileptogenic network by analyzing dynamic metabolic parameters. Such a multidimensional assessment method provides new insights into the pathophysiological mechanisms of epilepsy and creates conditions for developing personalized treatment plans.

Although dual-probe PET/MR has certain application potential, the current clinical use of PET/MR for MRI-negative epilepsy lacks an operable standardized protocol, and physicians face many difficulties in actual practice, such as MR scanning protocols, procedures, fusion strategies, quantitative analysis support, and image interpretation. By proposing a standardized scanning protocol based on dual-probe PET/MR, this paper not only verifies the effectiveness of PET/MR in enhancing the accuracy of epileptogenic focus localization in patients with refractory epilepsy, but also tries to establish a set of standardized operation procedures that can be promoted, and promotes the translation of the technology to the clinic. This protocol specifies key aspects such as the selection of radiotracers, the design of imaging sequence, and the strategy of image interpretation, which provides a referable operation specification for subsequent clinical application.

The dual-probe PET/MR protocol proposed in this study is methodologically feasible, but the following conditions need to be met in practice: First, equipment requirements: integrated PET/MRI equipment is required; second, radiotracer preparation: ¹¹C-FMZ needs to be prepared on-site using a cyclotron; third, scanning process compatibility: the scanning sessions for ¹⁸F-FDG and ¹¹C-FMZ need to be scheduled at least 24 h apart; fourth, personnel technical requirements: professional nuclear medicine personnel who have mastered both metabolic analysis using PET and neuroimaging interpretation of MRI are required. These practical challenges may affect the adoption of this protocol in different clinical settings, and further optimization is needed to lower the implementation threshold in the future.

The study has received approval from the local medical ethics committee. This study adheres to the principles of the Declaration of Helsinki. Patients or their legal guardians must sign a written informed consent form that details the invasive procedures, radiation exposure, and follow-up requirements.

1. Preparation and quality control of radiotracers

NOTE: It is essential to adhere to the established principles of biological occupational protection and radiological occupational protection. Additionally, it is important to follow the guidelines for the proper and compliant disposal of medical and radioactive waste during all operational procedures, so as to ensure the safety and well-being of all personnel involved.

1. Core process for the preparation of ¹⁸F-FDG
   NOTE: The synthesis is fully automated. Complete the following steps in under 60 min to ensure high radiochemical purity and sterility under aseptic conditions for clinical PET diagnostics.
   1. Fluorine-18 ion ([18F]F⁻) production: Bombard oxygen-18-enriched water (H₂¹⁸O) with protons in a cyclotron.
   2. Capture and activation: Pass the [¹⁸F]F⁻ solution through an anion exchange column (e.g., QMA). Then elute and activate the column using an acetonitrile solution containing potassium carbonate (K₂CO₃) and a phase transfer catalyst (e.g., Kryptofix 2.2.2, K222).
   3. Azeotropic dehydration: Add acetonitrile to the residue. Then heat the mixture and repeatedly evaporate it under a stream of inert gas until a dry residue is obtained, ensuring complete removal of moisture to yield a highly reactive anhydrous fluorine source.
   4. Nucleophilic fluorination labelling: Mix the dry fluorine source with the precursor (Man(OTf)₂-Ac) in anhydrous acetonitrile, heat the reaction mixture, and generate the [₁₈F]-labelled intermediate ([¹⁸F]FDM).
   5. Alkali hydrolysis deprotection: Add sodium hydroxide and heat the mixture to hydrolyze and remove the acetyl protecting group, yielding crude ¹⁸F-FDG.
   6. Neutralisation and purification: Neutralize the hydrolysis solution, then purify it by solid phase extraction using an alumina column (to remove free fluoride ions and catalysts) followed by a C₁₈ reverse-phase column (to remove hydrophobic impurities).
   7. Sterile filtration and aliquoting: Collect the purified solution by filtration through a 0.22 µm sterile filter and aliquot it as required.
   8. Rapid quality control and release: Perform critical quality control tests, including activity, radiochemical purity, pH, and complete the testing within a very short timeframe (typically <30 min). If the product meets all specifications, release it for PET imaging.
      NOTE: Radiochemical purity (RCP) is evaluated by analytical thin-layer chromatography (TLC) according to pharmacopoeial standards, which specify a minimum RCP of 90%, with typical values exceeding 97%.
2. Core process for the preparation of ¹¹C-FMZ
   NOTE: Due to the short half-life of ¹¹C (20 min), the entire process from radionuclide production to the final sterile product must be completed within 30 min to ensure sufficient activity for imaging.
   1. Production and conversion of ¹¹C-CO₂: Use a cyclotron to generate ¹¹C-CO₂ via the ¹⁴N(p,α)11C nuclear reaction, employing a 1% O₂/N₂ mixture as feedstock.
   2. Generate ¹¹C-trifluoromethanesulfonyl methyl ether (¹¹C-Triflate-CH₃): transfer ¹¹C-CO₂ to the drug synthesis system and further convert into ¹¹C-Triflate-CH₃.
   3. Neutralization with HCl: Dissolve 0.5 mg of demethylflumazenil and 1 mg of NaH in 0.4 mL of Dimethylformamide (DMF) within the reaction cell. React this mixture with ¹¹C-triflate-CH₃ at 0 °C for 5 min, then bring it to room temperature (RT) and neutralize it with 0.5 mol/L HCl.
   4. High-performance liquid chromatography (HPLC) separation and purification: Separate and purify the mixture via HPLC. Separation column: Nucleosil 100-5 C₁₈, 250 mm × 10 mm. Mobile phase: 22% acetonitrile solution containing 0.01 mol/L phosphoric acid. Detection wavelength: 254 nm. Flow rate: 4 mL/min.
   5. Collecting the radioactive peak fraction: Perform additional purification using a C₁₈ column, followed by final filtration through a 0.22 µm sterile filter membrane to obtain the final product, ¹¹C-FMZ.
   6. Synthesis quality verification of ¹¹C-FMZ: Verify the radiochemical purity (RCP) by performing radiochemical purity analysis of the synthetic product via HPLC. Chromatographic conditions are as follows: Column: Reverse-phase C₁₈ column; Mobile phase: Methanol/water = 65/35 (v/v); Flow rate: 1.0 mL/min; Detector: Radioactivity detector for online monitoring of radioactive signals. Ensure that the retention time of the radioactive peak corresponding to the target compound (¹¹C-FMZ) matches that of the known standard (typically 6.7 min).
      NOTE: Radiochemical purity is defined as the percentage of the radioactive integral area of the target peak relative to the total radioactive peak area. A value ≥98 % is required to deem the batch qualified and release it. A successful HPLC purification is indicated by a sharp, symmetric radioactive peak on the chromatogram at the expected retention time for ¹¹C-FMZ, with good separation from earlier or later peaks. The collected fraction should appear as a clear, colorless solution. However, its essence lies in verifying the preparation results, falling under In-Process Control (IPC) rather than comprehensive quality testing for final product release (such as sterility and endotoxin testing).

2. Patient inclusion criteria

NOTE: This study aims to establish a standardized dual-probe PET/MR imaging protocol. Sample size selection is based on feasibility rather than statistical power calculations, as the primary objective is to establish procedural reproducibility and preliminary technical feasibility within a well-defined patient population.

1. Apply the following criteria based on the International League Against Epilepsy (ILAE) consensus²⁵ definition:
   1. Document the failure of two prior anti-epileptic drug (AED) regimens.
   2. Confirm that each AED trial was adequate in terms of drug selection (appropriate for the seizure type or epilepsy syndrome), dosage (titrated to a tolerated, clinically effective dose), and duration (sufficient to assess efficacy, typically at least three months).
   3. Define treatment failure as the recurrence of any seizure type (including auras) after an adequate trial, thereby failing to achieve sustained seizure freedom for a minimum duration of 12 months.
2. Auxiliary diagnosis for patients with refractory epilepsy.
   Synthesize the following evidence: epileptiform discharges on electroencephalogram (EEG) (spikes, spike-slow wave complexes), structural lesions on MRI (e.g., hippocampal sclerosis), and progressive cognitive or behavioral impairments revealed by neuropsychological evaluations in order to support a diagnosis of refractory epilepsy and to suggest its functional consequences.
3. Exclude ineligible subjects based on the following criteria:
   1. Exclude if absolute contraindications include recent (within 3 months) major surgery, trauma, or infection.
   2. Exclude other clinical trial participants who have not gone through a washout period.
   3. Exclude women who are pregnant or breastfeeding.
   4. Exclude those who are claustrophobic or unable to cooperate with prolonged scanning.
   5. Exclude patients with the presence of contraindications to PET/MR devices (e.g., metal implants, pacemakers). Relative contraindications include poor glycemic control (fasting glucose >11.1 mmol/L), severe hepatic or renal failure that may affect the metabolism of the imaging agent, and severe cardiopulmonary or metabolic disease (e.g., diabetes). In addition, the investigator will need to exclude patients with severe depression, anxiety, or psychotic symptoms to ensure their cooperation with the study.

3. Standardized protocol for dual-probe PET/MR imaging

1. Perform the scanning according to the following procedure:
2. Perform an FDG PET/MR scan to obtain a whole-brain metabolic profile. Subsequently, perform FMZ PET/MR scanning at least 24 h later, designed to avoid residual FDG interfering with the quantitative analysis of the FMZ.
   ​NOTE: The extensive hypometabolic areas revealed by FDG can provide a key reference for the subsequent precise boundary outlining and lesion localization of the FMZ. See Figure 1.

Figure 1: Inspection protocol flow chart. Schematic diagram of the dual-probe PET/MR (¹⁸F-FDG/¹¹C-FMZ) scanning protocol. Please click here to view a larger version of this figure.

4. Preparation for PET/MR examination

1. Within 24 h prior to the study, instruct participants to avoid alcohol, caffeine, non-essential medication, and strenuous exercise. Confirm compliance before scanning to maintain stable physiological conditions (e.g., biological rhythms and cognitive function) and ensure data accuracy.
   NOTE: Patients need to be in the inter-seizure period. Patients should remain seizure-free until 24 h prior to receiving ¹⁸F-FDG and ¹¹C-FMZ PET/MR scans²⁶. This timeframe helps ensure that observed metabolic reduction reflects the interictal state rather than transient postictal changes, thereby enhancing specificity in localizing the epileptogenic focus. The same principle applies to ¹¹C-FMZ imaging to avoid alterations in GABA_A receptor availability surrounding the ictal period.
2. Management of AEDs
   1. Prior to the 11C-FMZ PET scan, withhold AEDs that potentially interfere with GABA_A receptor binding (e.g., benzodiazepines, barbiturates) for a period sufficient to eliminate their pharmacological effects, typically at least 5 half-lives. However, in patients who are at significantly higher risk of seizures after discontinuation, perform adjustment of their AEDs under the close assessment and supervision of the attending epilepsy specialist. The core principle is to prioritize patient safety.
   2. Risk assessment and decision-making: Determine whether and how to adjust AEDs (e.g., dose reduction rather than complete discontinuation) based on the patient's comprehensive assessment, including their daily seizure frequency, severity, and medical history, and have a clinician sign the written protocol.
   3. Alternative protocols: If the risk of discontinuing the relevant AEDs is deemed too high by the clinician, perform scanning without adjusting the drug regimen. However, document the name of the medication taken, the dosage, and the time of the last dose clearly in the image report so that the possible competitive inhibitory effect of the corresponding medication on GABA_A receptor binding can be taken into account when interpreting the ¹¹C-FMZ images.
   4. Treatment of deviations from the protocol: Consider any deviation from the ideal drug administration protocol a scanning failure, but rather a reflection of a real clinical scenario. Analyze and document the interpretation of the images in the context of the specific medication situation.
3. Prior to FDG PET/MR scanning, instruct participants to fast for 8 h, avoiding food, sugary beverages, and intravenous glucose. Verify blood glucose levels remain below 11.1 mmol/L before imaging.
   NOTE: Before the FMZ PET/MR examination, normal unsweetened plain water may be consumed prior to the examination. If the patient needs to eat, it should be done at least 2 h before the scan and limited to light foods (e.g., white porridge, white bread) that do not contain caffeine or alcohol.
4. Review the application form to verify the subject's basic information, the purpose of the examination, and the program; ensure that all preparatory work has been completed; and record the subject's details in the file.
5. Obtain informed consent for PET/MR imaging from subjects and their families, detailing the purpose of the examination, the procedure, the potential risks, and the expected benefits; as well as reviewing the medical history in detail, recording height and weight, and confirming that there were no contraindications to MRI.
6. Establish peripheral superficial venous access, and double-check the subject's information, type of imaging agent, and dose before injection. Inject the contrast agent slowly intravenously, followed by flushing the tubes with an appropriate amount of saline to minimize residuals. After injection, use gauze to press the puncture point to prevent leakage, and record the injection time, site, and activity accurately.
   NOTE: The specific administered activity for each patient is standardized based on body weight. The dose is calculated using the formula: Administered Activity (MBq) = Patient's Body Weight (kg) × Standard Dose per Kilogram (MBq/kg). To ensure consistency, a fixed value within the specified range is selected for all patients in our protocol (FDG: 5 MBq/kg, FMZ: 4.5 MBq/kg). The calculated dose is accurately drawn into a syringe and verified by a second technologist before injection.
7. Injections and timing of uptake
   1. Injection and timing of uptake for ¹¹C-FMZ.
      NOTE: Due to its extremely short physical half-life (20 min), the entire process from injection to the start of the scan must be efficient and precise.
      1. Complete Intravenous (IV) injection in a dedicated preparation room. Immediately after the injection, escort the patient to the PET/MR scanning room and prepare the patient for positioning, posing, and coil mounting.
      2. Time control: Limit this post-injection preparation process to less than 20 min, with the goal of starting PET data acquisition as soon as possible. Any unnecessary delay will result in significant attenuation of the tracer activity, which will severely affect the signal-to-noise ratio and diagnostic value of the images.
      3. Scan Initiation: Perform PET acquisition within a time window of 20 min after injection. This time window corresponds to a quasi-equilibrium state of FMZ binding to the receptor, with the best signal-to-noise ratio.
   2. Injection and timing of uptake for 18F-FDG
      NOTE: Timing of injection is relatively flexible, but adequate uptake time needs to be ensured.
      1. Perform intravenous injection in a quiet, light-proof preparation room. After injection, keep the patient in a resting state for 30 min to ensure that the brain's uptake of FDG has reached homeostasis, followed by a trip to the PET/MR scanning room and preparations such as positioning, posing, and coil mounting.
      2. Scan initiation: Initiate PET acquisition around 40 min after injection, a time window that corresponds to a quasi-equilibrium state of FDG binding to the receptor and optimal signal-to-noise ratio.
8. After completing the injection, immediately initiate the audiovisual isolation procedure. Dim the lights in the waiting room and set the temperature to approximately 22 °C. Instruct the subject to close their eyes, rest quietly in bed, and avoid talking or eating during this period.
   NOTE: This standardized resting state is critical for ensuring consistent cerebral uptake and distribution of the radiotracers, especially for ¹⁸F-FDG, which is highly sensitive to neural activity and environmental stimuli.

5. PET/MR imaging process

1. Before the examination, instruct the subject and accompanying persons to remove all metal objects, including mobile phones, headwear, dentures, glasses, watches, wallets, and coins. Additionally, prohibit the entry of wheelchairs, stretchers, examination beds, oxygen cylinders, or monitoring equipment into the scanning room.
2. Provide the subject with earplugs and position them supine on the PET/MR examination table, and then use a head-neck coil (8-channel head-neck combined coil) to encircle the neck area and secure the head with an immobilization device (e.g., specialized headrest or foam pad) to minimize motion while ensuring comfort.
3. Instruct the subject to keep their arms relaxed at their sides, and inform them to activate the alarm device if they feel uncomfortable.
4. Review the acquired images to confirm the subject's head is correctly centered within the scanner and aligned with the coil center.
5. Acquire PET data using the Ordered Subsets Expectation Maximization (OSEM) algorithm for image reconstruction. Perform MR imaging using the Zero Echo Time (ZTE) pulse sequence for attenuation correction, which enables segmentation of bone, air, and soft tissue.
   NOTE: According to the results of the neurological function assessment, appropriate doses of sedative medications should be administered to the patient, if necessary, to keep them comfortable and stable during the PET/MR exam. This helps ensure the quality of the images and prevents motion blur caused by the patient's discomfort or anxiety. Such motion blur could impact the clarity and accuracy of the final images.

6. PET/MR scanning sequences and imaging strategies

NOTE: The PET/MR system used is the GE Healthcare SIGNA PET/MR (3.0 T MR with LBS-SiPM detector).

1. Perform MR and PET scans synchronously, with a PET scan time of 40 min. MR sequences are divided into structural and functional sequences. Under normal circumstances, only one functional image is acquired during the two acquisition sessions. The decision depends on the patient's condition during these sessions; if the condition is better, allow an appropriate increase in MR scanning sequences. For comparisons between seizure episodes and interictal periods, selectively acquire twice functional images. See Table 1.
   ​NOTE: The role of MR functional sequences is auxiliary and supplementary. Core epileptogenic focus localization still relies on the fusion analysis of dual-probe PET and high-resolution structural MRI. Researchers may select the approach based on specific circumstances.

Table 1: Acquisition objectives and parameters for each specific sequence. Please click here to download this Table.

7. Interpretation of results

1. Image quality control and preprocessing
   1. Perform an assessment of original image quality, including preliminary quality control for structural MRI, functional imaging, ¹⁸F-FDG PET metabolic imaging, and ¹¹C-FMZ PET receptor binding imaging. Proceed with only images passing the aforementioned quality control checks to subsequent visual and quantitative analysis workflows. Process or discard images failing quality control according to the protocol outlined in Section 8.
      NOTE: This step was accomplished using image analysis software on a GE Advantage Workstation 4.7 workstation, specifically Volume Viewer 7 software for automated fusion. Measurement accuracy errors of 0.1 mm and 0.1 degrees minimum, and the fusion error is a voxel.
   2. Fuse ¹⁸F-FDG PET metabolic maps and ¹¹C-FMZ PET receptor binding maps with high-resolution 3D T1-weighted anatomical images, with T2-FLAIR sequence images displayed side-by-side as key references, enabling preliminary visual assessment of lesion extent.
      NOTE: Fusion is considered successful when the PET metabolic/receptor binding images and T1-weighted MRI anatomical images achieve perfect registration of key structures -- including cortical anatomical contours, ventricular boundaries, and the hippocampus -- with no ghosting or edge displacement.
2. Quantitative analysis
   1. Measure the standardized uptake values (SUV) of lesions. Lesions associated with epilepsy often exhibit SUV values lower than the contralateral side or the healthy control group. This interpretation requires comparison with the contralateral normal brain region or a healthy control database, rather than using absolute thresholds.
   2. Calculate the asymmetry index (AI) value. In clinical practice, an AI value >10-15% is typically used as the threshold for identifying metabolic abnormalities in lesions. The formula is: AI(%) = |(Contralateral SUV mean - Lesion SUV mean)| / Contralateral SUV mean × 100%.
   3. Assess the spatial relationship between FDG abnormal regions and FMZ abnormal regions based on region of interest (ROI), categorizing overlap into three types: complete overlap (the anomalous regions exhibit high spatial consistency), partial overlap (the FMZ abnormality zone is entirely contained within the FDG abnormality zone, but it is smaller and more localized), or complete separation (anomalous regions are spatially independent of each other).
      NOTE: Institutions are strongly advised to establish internal laboratory standards based on their specific equipment, reconstruction algorithms, and segmentation methods when applying this protocol to achieve the most accurate results. All quantitative analyses should support visual interpretation, primarily serving to provide objective quantification for visually identified suspicious abnormalities and to offer supplementary evidence for decision-making when visual assessment is inconclusive.
3. Systematic analysis of dual-probe images
   1. Perform independent analysis of FDG PET images.
      1. Identify regions of abnormal glucose metabolism, focusing on the anatomical location, spatial extent, intensity (quantified by SUV), and morphological characteristics (e.g., boundary clarity) of hypometabolic areas.
      2. Combine visual analysis with quantitative metrics, paying particular attention to focal hypometabolic areas in the cerebral cortex and deep structures.
   2. Perform independent analysis of FMZ PET images.
      1. Evaluate GABA_A receptor distribution. Identify regions with abnormal radiotracer binding (typically manifesting as reduced local binding density).
      2. Focus on their anatomical location, spatial extent, and whether they exhibit sharper boundaries compared to FDG hypometabolic areas.
   3. Integrate dual-tracer findings:
      1. Prioritize regions showing concurrent ¹⁸F-FDG metabolic reduction and ¹¹C-FMZ binding decrease as a high-probability epileptogenic focus. This process is based on ROI and quantitative metrics, displaying FDG and FMZ images side-by-side for a comprehensive assessment of results.
   4. Integrate multimodal imaging findings through the following approaches:
      1. Correlate anatomically identified regions from FDG and FMZ PET scans.
      2. Assess spatial concordance.
      3. Quantitatively analyze imaging parameter differences.
      4. Make comprehensive judgments incorporating clinical and electrophysiological data. Localize epileptic foci to specific lobes, gyri, or subcortical structures.

8. Troubleshooting and quality assurance

NOTE: This protocol is designed to ensure the high quality of the data collected. The following summarizes common problems, their potential causes, and recommended adjustments. Before implementing any adjustments, the impact on patient safety, radiation dose, and integrity of diagnostic information should be weighed.

1. Poor image fusion
   1. Prevention: Use a repeatable head immobilization device for each scan to ensure consistent body position.
   2. Treatment: Perform manual or semi-automatic fine fusion using workstation software. If there are fusion errors, record these and consider their effect in the interpretation of the results. If they cannot be satisfactorily corrected, consider the study invalid.
   3. Review: Always visually verify fusion accuracy by overlaying the display and checking the fusion of key anatomical structures (e.g., hippocampus, ventricular margins).
2. Motion artifacts
   1. Prevention: Communicate adequately before scanning to obtain patient cooperation. For anxious or uncooperative patients, use sedatives as appropriate according to preset criteria.
   2. Recognition: Monitor the image reconstruction process real-time, and immediately pause the scan when artifacts are detected.
   3. Treatment of artifacts.
      1. For mild artifacts, attempt image reconstruction using motion correction algorithms.
      2. For severe artifacts, consider reacquisition of the affected portion of the sequence (especially for MR sequences) if the patient's status permits and tracer activity is sufficient.
      3. For FMZ scans, quickly assess whether the remaining activity supports reacquisition.
3. ¹¹C-FMZ scanning delay
   1. Prevention: Optimize synthesis and quality control processes to ensure a seamless transition. Complete all non-invasive preparations prior to injection.
   2. Contingency:
      1. For a short delay (<10 min), perform immediate collection as originally planned. Realize that attenuation of activity may result in reduced signal-to-noise ratio of the image, and this needs to be noted in the report.
      2. For longer delays (>10 min), adjust the acquisition time. Follow stepa 8.2.3.2-8.3.2.5.
      3. Defer the acquisition start point, but ensure that the acquisition duration is sufficient (e.g., 30-50 min post-injection). This may reduce quantitative accuracy, but may salvage the qualitative diagnosis.
      4. Adjust reconstruction parameters, use algorithms or parameters that target low count rates during image reconstruction to optimize image quality.
      5. Rescanning: If the delay is too long, resulting in a significant lack of activity and conditions permit (e.g., hospital policy, ethical approval, patient consent), consider rescanning with a freshly prepared tracer on the following day or another day.

All case images in this study were subjected to a strict quality control process as specified in paragraph 7.1 and section 8 of this protocol. They were included in the analysis only after ensuring that they met the following criteria: (1) the MRI images had no obvious motion artifacts or geometric distortion; (2) the tracer distribution of PET images was uniform, and the signal-to-noise ratio met the requirements for quantitative analysis; and (3) the automated fusion of the PET and MRI images was successful, and it was confirmed by visual inspection that the key anatomical structures (e.g., hippocampus and sulcus gyrus contours) were aligned without error, and that there was no fusion error that would affect the diagnosis. All the representative images presented met the above quality control standards.

Patient A, male, 28 years old
A decade ago, the patient experienced the initial onset of seizures, characterized by a 10-s absence episode involving loss of consciousness and incoherent speech, without limb stiffness. Initial hospital evaluations at that time were unremarkable. Subsequently, the seizure frequency increased and the semiology evolved, incorporating symptoms such as postictal amnesia and longer durations of up to 30 s, accompanied by automatisms including drooling and lip-smacking. Over the following 2 years, the seizures further progressed to include bilateral convulsive activity. The typical presentation involved loss of consciousness, version of the head and eyes, foaming at the mouth, and bilateral limb jerking, with each episode lasting approximately 30 s. These were followed by a prolonged postictal drowsiness period of about 30 min. The condition escalated significantly, with the patient experiencing daily clusters of seizures over a seven-day period, exceeding ten episodes per day. This culminated in a hospital admission in 2023 due to this heightened seizure activity, accompanied by a notable decline in memory function.

Treatment history: The patient's initial workup at Hospital 1, including an EEG and head MRI, revealed abnormal brain activity. An initial trial of Chinese herbal medicine provided no clinical benefit. Subsequent treatment with oxcarbazepine at Hospital 2 failed to reduce seizure frequency. After a formal diagnosis of epilepsy was established at Hospital 3, therapy was switched to carbamazepine, which also failed to achieve adequate seizure control.

PET/MR results: No structural abnormalities were detected on T1-weighted MRI. T2-weighted imaging showed a slight increase in signal intensity in the right hippocampus, with lower SUV (FDG: 5.2; FMZ: 2.3) compared to the contralateral side (FDG: 7.3; FMZ: 4.7). The asymmetry indices (AI) are all greater than 15 % (FDG: 29 %; FMZ: 51 %). The abnormal regions detected by FMZ and FDG PET show a high degree of spatial consistency. See Figure 2 and Figure 3.

Figure 2: FDG PET/MR images of Patient A during the interictal period. (A-C) The raw PET images in the coronal, sagittal, and transverse planes. (D-F) The raw T1-weighted MRI images (T1WI) in the coronal, sagittal, and transverse planes. (G-I) The fused images of PET and T1WI in the coronal, sagittal, and transverse planes. (J-L) Throw the raw T2 FLAIR images in the coronal, sagittal, and transverse planes. No structural abnormalities were detected on T1-weighted MRI. The T2 FLAIR images demonstrated slightly elevated signal intensity, with the right hippocampus exhibiting a lower standardized uptake value (5.2) compared to the contralateral side (7.3), yielding an asymmetry index of 29%. Please click here to view a larger version of this figure.

Figure 3: FMZ PET/MR images of Patient A during the interictal period. (A-C) The raw PET images in the coronal, sagittal, and transverse planes. (D-F) The raw T1-weighted MRI images (T1WI) in the coronal, sagittal, and transverse planes. (G-I) The fused images of PET and T1WI in the coronal, sagittal, and transverse planes. (J-L) The raw T2 FLAIR images in the coronal, sagittal, and transverse planes. No structural abnormalities were detected on T1-weighted MRI. The T2 FLAIR images demonstrated slightly elevated signal intensity, with the right hippocampus exhibiting a lower standardized uptake value (2.3) compared to the contralateral side (4.7), yielding an asymmetry index of 51%. Please click here to view a larger version of this figure.

Patient B, female, 17 years old
Clinical records indicate the patient's initial seizure occurred 12 years ago, characterized by a sudden, unprovoked loss of consciousness. The event involved limb convulsions, upward eye deviation, and oral frothing, lasting several seconds. The patient experienced full recovery but had no memory of the incident (postictal amnesia). She received a diagnosis of "epilepsy" from a local neurology hospital and was started on oral AEDs, though the specific agents are undocumented. Due to repeated self-discontinuation of the prescribed medication, the patient's seizure control was suboptimal. Over the past 4 years, the seizure frequency has increased to at least once per week.

Medical history: In early 2025, the patient was prescribed oxcarbazepine (300 mg/day). Following the initiation of therapy, the patient developed a generalized rash, prompting discontinuation of the drug due to a suspected allergic reaction. On April 2, 2025, the patient experienced two generalized tonic-clonic seizures, leading to a hospital admission for further evaluation and management.

PET/MR results: No structural abnormalities were detected on T1-weighted MRI. T2-weighted imaging showed a slight increase in signal intensity in the left hippocampus, with lower SUV (FDG: 5.9; FMZ: 2.8) compared to the contralateral side (FDG: 7.9; FMZ: 5.2). The asymmetry indices (AI) are all greater than 15 % (FDG: 25 %; FMZ: 46 %). The FMZ abnormality zone is located within the FDG abnormality zone, with a smaller extent. See Figure 4 and Figure 5.

Figure 4: FDG PET/MR images of Patient B during the interictal period. (A-C) The raw PET images in the coronal, sagittal, and transverse planes. (D-F) The raw T1-weighted MRI images (T1WI) in the coronal, sagittal, and transverse planes. (G-I) The fused images of PET and T1WI in the coronal, sagittal, and transverse planes. (J-L) The raw T2 FLAIR images in the coronal, sagittal, and transverse planes. No structural abnormalities were detected on T1-weighted MRI. The T2 FLAIR images demonstrated slightly elevated signal intensity, with the left hippocampus exhibiting a lower standardized uptake value (5.9) compared to the contralateral side (7.9), yielding an asymmetry index of 25%. Please click here to view a larger version of this figure.

Figure 5: FMZ PET/MR images of Patient B during the interictal period. (A-C) The raw PET images in the coronal, sagittal, and transverse planes. (D-F) The raw T1-weighted MRI images (T1WI) in the coronal, sagittal, and transverse planes. (G-I) The fused images of PET and T1WI in the coronal, sagittal, and transverse planes. (J-L) The raw T2 FLAIR images in the coronal, sagittal, and transverse planes. No structural abnormalities were detected on T1-weighted MRI. The T2 FLAIR images demonstrated slightly elevated signal intensity, with the left hippocampus exhibiting a lower standardized uptake value (2.8) compared to the contralateral side (5.2), yielding an asymmetry index of 46%. Please click here to view a larger version of this figure.

Representative case selection basis and method validation description
The representative cases described in this paper are intended to validate this dual-probe PET/MR protocol in the following three ways:

Reproducibility validation: Case A and Case B were different in gender, age, and disease duration, but both completed the scanning and image analysis in strict adherence to a standardized procedure as described in Part 3 of this protocol. This demonstrates the ability of this protocol to be implemented reproducibly in different individuals.

Robustness validation: The two patients presented different imaging patterns. The FDG hypometabolic zone was highly consistent with the FMZ hypobinding zone in case A; the extent of FMZ abnormality was significantly smaller than that of the FDG abnormality zone in case B. This difference reflects the complexity of the epilepsy pathology network, and the ability of this protocol to clearly capture and rationally interpret these two patterns demonstrates its adaptability and robustness to different pathophysiological features.

Diagnostic value validation: Based on the localization results of this protocol, both patients underwent surgical treatment and achieved the desired outcome of postoperative seizure-free epilepsy. This outcome confirms the accuracy of the localization results of this protocol and its clinical diagnostic value.

The primary objective of this study was to establish and delineate a standardized operational protocol for dual-probe (¹⁸F-FDG/¹¹C-FMZ) PET/MR in the evaluation of MRI-negative refractory epilepsy. The pressing clinical need for such a protocol stems from the current lack of uniform specifications across centers, which leads to inconsistencies in imaging acquisition, analysis, and interpretation, thereby hindering the comparability of results and broader clinical adoption. This protocol aims to address this gap by providing a detailed, step-by-step framework covering key aspects such as radiotracer preparation, patient preparation, imaging sequence design, and a dual-tracer image interpretation strategy.

The study presents two representative cases. In Patient A, congruent hypometabolism (FDG) and reduced receptor binding (FMZ) colocalized to the right hippocampus. Patient B, in contrast, exhibited extensive left hippocampal hypometabolism that encompassed a more focal region of reduced FMZ binding. These findings suggest that the combined modality can delineate distinct pathophysiological processes. FDG hypometabolism reflects a broad zone of low neuronal activity due to synaptic inhibition or dysfunction^19,27, reduced FMZ binding provides a more direct marker of neuronal loss or GABA_A receptor downregulation²⁸. In Patient B, the FMZ abnormality was confined to a region smaller than the FDG hypometabolic area, thus pinpointing the likely core of the epileptogenic focus, which is surrounded by a larger area of metabolic impairment. This difference in spatial distribution highlights how adding ¹¹C-FMZ improves specificity by accurately depicting epileptogenic core focus within a broader metabolic network, a finding supported by previous studies²⁹.

After injection of ¹⁸F-FDG, it takes 40-60 min to reach metabolic steady state. Areas of low metabolism are extensive, allowing preliminary screening of suspected epileptogenic focus. During the interictal period of epilepsy, glucose metabolism decreases due to neuronal functional damage. FDG can sensitively detect these widespread hypometabolic regions; however, it has low specificity because the FDG signal boundaries are blurred and often extend beyond the actual epileptogenic focus^13,30. To achieve more precise delineation, FMZ imaging is subsequently employed. The half-life of ¹¹C-FMZ is only 20 min, with uptake peaking 10-20 min post-injection. Based on the regions identified by FDG, FMZ specifically targets suspected epileptogenic focus for imaging, providing clear boundaries of receptor binding deficiency and effectively overcoming the blurred range limitation of FDG. FMZ can precisely display receptor-deficient regions with clear boundaries, exhibiting significantly higher specificity than FDG. It is particularly valuable in localizing focal cortical dysplasia (FCD) not detected by MRI and temporal lobe epilepsy¹⁴. Additionally, the spatial correlation between metabolic activity and receptor expression enhances localization accuracy. Regions of low metabolism identified by FDG often surround the core areas of reduced receptor binding detected by FMZ imaging. For example, in cases involving hippocampal pathology, FMZ can precisely delineate lesions within the hippocampus.

Conventional MRI excels in anatomical resolution but has a high false-negative rate for lesions without structural abnormalities³¹. In contrast, SEEG (intracranial electrodes) is considered the physiological 'gold standard' but is invasive, costly, and has limited coverage^32,33,34. Single-probe PET excels in functional sensitivity, but has relatively vague anatomical localization^35,36. Dual-probe PET/MR combines the advantages of non-invasive, multidimensional, and precise localization. This 'metabolic broad-range screening and receptor-targeted precision' dual-modality logic, based on the dual-probe PET/MR protocol, has improved the accuracy of epileptogenic focus localization to over 90%; this is significantly higher than that of MRI or single-probe PET alone. More importantly, the dual probe PET/MR scanning feature radically improves diagnostic efficiency. It reduces the preoperative evaluation cycle and minimizes intermodal fusion errors between images acquired at different times by integrating independent MRI, FDG PET examinations, and potentially invasive monitoring, which traditionally require multiple visits, into one streamlined process³⁷. In addition, the dual-probe technique allows for flexible adjustment of the FMZ scanning range and protocol, reducing patient discomfort through segmented scanning. This strategy fundamentally solves the localization blind spot of traditional MRI in MRI-negative epilepsy, providing objective evidence at the molecular pathological level for determining the surgical resection range.

This protocol integrates multiple MR functional sequences, such as diffusion tensor imaging (DTI), arterial spin labelling (ASL), and susceptibility-weighted imaging (SWI). These synchronously acquired MR functional data provide additional physiological and structural dimensions for assessment. For example, DTI can reveal damage to the integrity and abnormal connections of white matter fibre tracts associated with epileptogenic focus^38,39,40. ASL can reflect changes in local cerebral blood flow perfusion, which may be associated with metabolic abnormalities⁴¹. Additionally, SWI can help identify potential epileptogenic structural abnormalities, such as microhaemorrhages or vascular malformations. Future studies could further explore the quantitative correlations between these multimodal functional parameters, for example, PET metabolism and receptors, as well as MR DTI, ASL, and SWI, to construct a more comprehensive "metabolism-receptor-structure-connection" epilepsy network model. This approach would provide richer imaging evidence for understanding the complexity of epilepsy networks and selecting individualised neurostimulation targets.

Although this protocol significantly improves localization accuracy, it still faces challenges related to probe preparation and cost. The short half-life of ¹¹C-FMZ, only 20 min, necessitates on-site synthesis via a cyclotron, limiting its adoption in primary care hospitals. Additionally, the method is highly dependent on equipment, with a limited number of PET/MR devices available and relatively high examination costs. Nevertheless, dual-probe PET/MR (FDG+FMZ) achieves non-invasive, precise localization of epileptogenic focus in refractory epilepsy by employing multimodal imaging. To accelerate future clinical translation, developing and adopting GABA_A receptor probes with longer half-lives, such as ¹⁸F-FMZ, will be key to overcoming the reliance of ¹¹C-FMZ on on-site cyclotrons. This will significantly advance the clinical translation of this protocol across a broader range of medical institutions.

In conclusion, this study aims to enhance the usability of dual-probe PET/MR technology in clinical practice by providing a set of standardized operational protocols. Currently, the clinical application of this technique may be somewhat limited by the lack of uniform specifications, with variations in specific operational details across centers, which may affect the consistency and comparability of results. This protocol attempts to provide an initial frame of reference for operators by refining key steps, such as clarifying the interval between two scans, the time frame for operation after tracer injection, and the basic criteria for image fusion. We expect that this standardization effort will reduce the difficulty of operation to a certain extent, facilitate the conduct of similar examinations in different medical centers and the exchange of results, and thus may contribute to the exploration of a wider range of future applications of this technology.