Proposal for a magnetic resonance imaging protocol for the detection of epileptogenic lesions at early outpatient stages


  • Jörg Wellmer,

    Corresponding author
    1. Ruhr-Epileptology, Department of Neurology, University Hospital Knappschaftskrankenhaus, Bochum, Germany
    • Address correspondence to Jörg Wellmer, Ruhr-Epileptology, Department of Neurology, University Hospital Knappschaftskrankenhaus, In der Schornau 23-25, 44892 Bochum, Germany. E-mail:

    Search for more papers by this author
    • Members of the ad hoc commission “Structural Imaging” of the German Section of the International League Against Epilepsy (ILAE).

  • Carlos M. Quesada,

    1. Department of Epileptology & Life and Brain Institute, University Hospital Bonn, Bonn, Germany
    Search for more papers by this author
  • Lars Rothe,

    1. Ruhr-Epileptology, Department of Neurology, University Hospital Knappschaftskrankenhaus, Bochum, Germany
    Search for more papers by this author
  • Christian E. Elger,

    1. Department of Epileptology & Life and Brain Institute, University Hospital Bonn, Bonn, Germany
    Search for more papers by this author
  • Christian G. Bien,

    1. Epilepsy Center Bethel, Krankenhaus Mara, Bielefeld, Germany
    Search for more papers by this author
  • Horst Urbach

    1. Department of Neuroradiology, University Hospital Freiburg, Freiburg, Germany
    Search for more papers by this author
    • Members of the ad hoc commission “Structural Imaging” of the German Section of the International League Against Epilepsy (ILAE).



Magnetic resonance imaging (MRI) is a key technology in the presurgical evaluation of patients with epilepsy. Already at early outpatient stages it can contribute to the identification of patients who are, in the case of pharmacoresistance, good candidates for epilepsy surgery. Yet, “standard head” MRI examinations often fail to displaying therapeutically relevant epileptogenic lesions. The purpose of this study is to identify an epilepsy-specific MRI protocol, which is likewise sensitive for even small epileptogenic lesions and economical enough to be applied outside specialized epilepsy centers.


Based on a large European presurgical epilepsy program comprising 2,740 patients we identified the spectrum of common epileptogenic lesions and determine the set of MRI sequences that are required for their reliable detection. Relying on a series of small, therapeutically particularly relevant lesions we determined the required slices thickness, slice angulations, and orientations for an epilepsy-specific MRI protocol.

Key Findings

Indispensable for early outpatient epilepsy specific MRI are fluid attenuated inversion recovery (FLAIR), T2-weighted, T1-weighted, and hemosiderin/calcification-sensitive sequences. Slice thickness for T2 and FLAIR must not exceed 3 mm. The T1 image should be acquired in three-dimensional technique at 1 mm isotropic voxels size. For T2 and FLAIR, at least two slice orientations each must be demanded in hippocampal angulation. We suggest no adaption to a clinical focus hypothesis. The resulting “essential 6” sequence protocol allows the detection of virtually all common epileptogenic lesion entities.


The creation of a broadly accepted and abundantly applied MRI protocol for epilepsy outpatients can contribute to improved and earlier identification of potential candidates for epilepsy surgery. Our systematic analysis of MRI requirements for the detection of epileptogenic lesions can serve as basis for protocol negotiations between epileptologists, radiologists, and health care funders.

Epilepsy, in contrast to other chronic central nervous system disorders, has the particularity that in a subset of affected individuals, therapy can be performed that is directed toward the cause of the disease: epilepsy surgery. Following epilepsy surgery, patients with medically intractable epilepsy achieve long-term seizure freedom significantly more often than after continued medical treatment alone (Engel et al., 2012). Achievement of long-term relief from seizures improves individual quality of life (Mohammed et al., 2012), and several studies indicate that seizure freedom reduces the risk of morbidity and mortality in patients with epilepsy (Téllez-Zenteno et al., 2007; Bell et al., 2010). After successful epilepsy surgery, health care costs decline (Langfitt et al., 2007).

Magnetic resonance imaging (MRI) plays an important role in identifying patients who benefit from epilepsy surgery. Postsurgical seizure freedom is achieved significantly more often when a circumscribed, resectable epileptogenic lesion can be identified on MRI preoperatively than when a patient is rated nonlesional (Bien et al., 2009; Téllez-Zenteno et al., 2010). In addition, MRI allows identification of conditions beforehand that preclude any success of surgery (e.g., bihemispheric migration disorders).

Yet, whether epileptogenic lesions are detected on preoperative MRI depends on the quality of the applied MRI and its evaluation. In 2002, Von Oertzen et al. found that following “nonexpert” reports of “standard” MRI scans, 61% of epileptogenic lesions remained undetected. Failure rate was reduced to 50% when expert neuroradiologists who were aware of a clinical focus hypothesis evaluated the “standard” MRIs again; however, only the application of an epilepsy tailored MRI protocol plus reading by experienced neuroradiologists resulted in an acceptable failure rate of 9% (Von Oertzen et al., 2002). Other studies had similar results (McBride et al., 1998).

Since McBride's and Von Oertzen's studies, progress has been made in MRI technology. In most countries 1.5 and 3 Tesla MRI is widely available; compared to 0.5–1.0 Tesla scanners, these increase spatial resolution and optimize the signal-to-noise ratio. In addition, new sequences and evaluation techniques have been implemented: three dimensional fluid attenuated inversion recovery (3D-FLAIR; Lummel et al., 2011), susceptibility weighted imaging (SWI; Saini et al., 2009), diffusion weighted imaging (DWI; Chatzikonstantinou et al., 2011), periodically rotated overlapping parallel lines with enhanced reconstruction (PROPELLER) sequences which reduce the extent of motion artifacts in brain images of moving patients (Nyberg et al., 2012), and MRI postprocessing (Wagner et al., 2011). Yet the everyday experience of specialized epilepsy units is that many epileptogenic lesions still remain undetected on outpatient MRIs. This contributes to often-documented, delayed referral of good epilepsy surgery candidates to presurgical assessment (Engel et al., 2012).

Possible reasons for undetected epileptic lesions in standard outpatient MRI are the following: (1) insufficient clinical focus hypotheses from the referring neurologists; (2) “standard head” protocols not optimized for the spectrum of epileptogenic lesions; and (3) unfamiliarity with the spectrum of epileptogenic lesions. Because even the best focus hypothesis and most profound knowledge of epileptogenic lesions do not permit the detection of lesions when they are invisible on the MRI scan, the starting point for any improvement of outpatient MRI diagnostics should be defining an MRI protocol that is adjusted to common epileptogenic lesions.

In this study we consider the spectrum of epileptogenic lesion found in 2,740 cases of presurgical workup to identify which MRI sequences, slice thickness, slice orientations, and angulations are required to maximize the sensitivity of early outpatient MRI for epileptogenic lesions, while keeping economic aspects in mind.

Materials and Methods

Spectrum of epileptogenic lesions

The spectrum of lesions that is associated with pharmacoresistant epilepsies is identified from 2,740 patients who underwent presurgical workup at the University of Bonn Epilepsy Center between 1989 and 2009 (Bien et al., 2013). In 1,835 patients, histologic results served for the classification of the epileptogenic lesion entity; in 905, mostly not operated on, the etiology is based on MRI evaluation. Because in some patients MRI showed more lesions than were finally resected, the identification of the spectrum of epileptogenic lesions is based on 2,763 lesions. For match of MRI-based and histopathologic diagnosis see Bien et al. (2013). Several of the identified lesion entities were subject to earlier publications from Bonn (see Box 1), but none of the previous studies gave an overview of the MRI properties of the lesions.

Box 1.

The following studies published after 2000 describe the spectrum of lesion entities found during the presurgical workup in Bonn in detail: (1) globally: Bien et al. (2013), Urbach et al. (2004); (2) by brain region: temporomesial: Clusmann et al. (2004); temporal neocortical: Schramm et al. (2001); temporobasal: Schramm & Aliashkevich (2008); occipital: Binder et al. (2008); parietal: Binder et al. (2009); insula: von Lehe et al. (2009); cingulate gyrus: von Lehe et al. (2012); frontal: Kral et al. (2001) and Schramm et al. (2002); (3) by entity: Ammon's horn sclerosis and limbic encephalitis: Soeder et al. (2009); Rasmussen encephalitis: Bien et al. (2005); long-term epilepsy-associated tumors: Urbach (2008); dysembryoplastic neuroepithelial tumor DNT/DNET simple and complex variants: Campos et al. (2009); isomorphic subtype of astrocytoma: Schramm et al. (2004); angiocentric neuroepithelial tumor (ANET): Majores et al. (2007); ganglioglioma: Luyken et al. (2004); focal cortical dysplasia with balloon cells: Urbach et al. (2002); subcortical band heterotopias: Huppertz et al. (2008); (4) at 3 Tesla: Urbach (2012).

MRI characteristics of epileptogenic lesion entities

The identification of MRI characteristics of epileptogenic lesions is based on the studies referred to in Box 1, cross checked with other publications on MRI of the respective lesion entities (see Box 2), and personal experience (in particular HU, JW). In addition, we analyzed signal characteristics of particularly small or mild lesion manifestations that were collected from three different MRI archives (Institute of Radiology, University of Bonn, Germany; Life and Brain-Institute/Department of Epileptology, University of Bonn, Germany; Epileptological-Neuroradiological Unit, Ruhr-Epileptology, Bottrop/Bochum, Germany).

Box 2.

The following studies not relying on the Bonn series describe characteristics of epileptogenic lesions in MRI: (1) globally: Deblaere & Achten (2008), Woermann & Vollmar (2009), Jackson & Badawy (2011); (2) by entity: Ammon's horn sclerosis: Malmgren & Thom (2012); pial angiomatosis Sturge-Weber: Hu et al. (2008), Griffiths (1996), Adams et al. (2009); cavernomas: de Champfleur et al. (2011), de Souza et al. (2008); high grade gliomas: Cha (2009); hypothalamic hamartomas: Freeman et al. (2004); encephalitis: Demaerel et al. (2011); DNT/DNET: Chassoux et al. (2012), Ozlen et al. (2010); oligodendroglioma: Khalid et al. (2012); pilocytic astrocytoma: Lee et al. (2011); meningeoma: Gasparetto et al. (2007); ANET: Lellouch-Tubiana et al. (2005); pleomorphic xanthoastrocytoma: Gocalves et al. (2013); supratentorial low grade glioma: Scarabino et al. (2005); epidermoid: Kallmes et al. (1997); Malformations of cortical development: Raymond et al. (1995); periventricular nodular heterotopias: Tassi et al. (2005); polymicrogyria: Barkovich (2010), Leventer et al. (2010); focal cortical dysplasia: Mellerio et al. (2012), Colombo et al. (2012); lissencephaly and subcortical band heterotopias: Dobyns (2010); (3) at 3 Tesla: Phal et al. (2008), Craven et al. (2012).

Lesion entities were evaluated for the following aspects

Suitability of MRI sequences for lesion entities

How suitable are commonly applied and abundantly available MRI sequences for the detection of the various epileptogenic lesions? The sequences comprise: T1-weighted imaging (performed as three dimensional T1 sequence with isotropic 1 mm voxels; 3D-T1); T2-weighted imaging (T2) or short tau inversion recovery (STIR), which can be used alternatively; fluid attenuated inversion recovery (FLAIR); T1-inversion recovery (IR); hemosiderin and calcification–sensitive sequences (hemo-calc); diffusion weighted imaging (DWI); and contrast-enhanced T1 (CM enh. T1). The suitability of sequences for lesion entities is rated according to the following scheme: “++” = sequence allows lesion detection even when lesions are subtle (high suitability); “+” = larger lesions may be detected, smaller may be missed (moderate suitability); “o” = even larger lesions can be missed (poor suitability).

We do not refer to lesion-specific particularities in 1.5 and 3 Tesla scanners, since the basic signal characteristics are similar. MRI at magnetic field strength below 1.5 Tesla is obsolete today for the identification of epileptogenic lesions because of poor contrast (see 'Discussion').

Required slice thickness

Because small epileptogenic lesions are usually those that provide the best chance for postoperative seizure freedom, MRI slice thickness should be adjusted to detect small lesions. Based on exemplary lesions we determine the minimal required slice thickness.

Recommendable slice orientation

Considering small lesions again, we analyze the need for multiple (coronal, axial, sagittal) cut planes to ensure that physiologic structures or partial volume effects within the complex three dimensionally folded cortex are not taken for pathology and vice versa.

Recommendable slice angulation

For T1 as well as coronal T2 and FLAIR, broad consensus exists on angulation. T1 images are often acquired in anterior commissure–posterior commissure (ac-pc) angulation because this facilitates anatomic orientation for radiologists who are not familiar with scans acquired along the hippocampal axis. Acquisition of the T1 sequence as isotropic 3D data set allows later reformatting into any angulation including the hippocampal axis following the given clinical question. Coronal T2 and FLAIR sequences have to be angulated perpendicular to the hippocampal (hc) axis to allow hippocampal volume estimation (important for the diagnosis of Ammon's horn sclerosis; Von Oertzen et al., 2002).

To evaluate which specific angulation recommendation should be given for axial T2 and axial FLAIR sequences, we evaluated the ease of detection of differently located lesions in 20 patients based on three dimensionally acquired 3 Tesla 1 × 1 × 1 mm T2 or FLAIR data sets, which we re-angulated into 3 mm slices along the hc and the ac-pc axis.

For hemo/calc sequences no angulation recommendation exists yet. The most important criterion should be minimization of disturbing susceptibility artifacts close to air-filled temporobasal and frontobasal sinuses. To identify the angulation with minimal brain loss due to susceptibility we performed a separate study on five volunteers who achieved axial SWI sequences at 1.5 and 3 Tesla in both hc and ac-pc angulation. These data will be published in detail elsewhere, but are summarized in this study.

Statistical analysis

To identify an economic protocol that relies on a minimum number of sequences but still allows detection of all epileptogenic lesion entities we performed a stepwise rank analysis. Based on Table 2, we identified the sequence that shows ++-suitability for most lesions. In a second step we identified the sequence that shows ++-suitability for most lesion types not already covered by the first ranked sequence. The same was repeated for the third and fourth ranked sequences (and so on).

The slice thickness and number of cut-plane orientations necessary to allow recognizing even small lesions are examined at exemplary cases without applying statistical measures. The superiority of hc- or ac-pc angulated axial T2/STIR and FLAIR was determined by consensus of three authors (JW, CMQ, and HU). The recommendation of hemo-calc angulation is based on the lower mean volume of brain loss due to susceptibility in hc or ac-pc angulated SWI scans.


Spectrum of epileptogenic lesions

The identified spectrum of epileptogenic lesions is shown in Table 1. Histologically separable lesion entities can share clinical and MRI characteristics. To facilitate the identification of necessary MRI sequences, these lesion entities are grouped (see legend of Table 1).

Table 1. Prevalence of epileptogenic lesions among 2,740 patients in tbe Bonn series 1989-2009
EntityN% of all% of lesional
  1. WHO, World Health Organization; n.o.s. = not otherwise specified.

  2. a

    Ganglioglioma; gangliocytoma; dysembryoplastic neuroepithelial tumor (simple and complex variant); supratentorial pilocytic astrocytoma; pleomorphic xanthoastrocytoma.

  3. b

    Angiocentric neuroepithelial tumor; isomorphic variant of astrocytoma; fibrillary astrocytoma; mixed low grade gliomas, low grade gliomas not further specified.

  4. c

    Anaplastic astrocytoma grade III; anaplastic ganglioglioma grade III; anaplastic oligodendroglioma and oligoastrocytoma grade III; anaplastic pleomorphic xanthoastrocytoma grade III; glioblastoma multiforme; gliosarcoma; gliomatosis cerebri; mixed glioma grade III.

  5. d

    Developmental venous anomaly (DVA) and arteriovenous malformation (AVM)— causality for epilepsy must be discussed individually.

  6. e

    Bonn series: exclusively meningioangiomatosis Sturge-Weber.

  7. f

    Including former diagnosis of glioneuronal hamartia/hamartoma and forme fruste of tuberous sclerosis (Wolf et al., 1997) now classified as focal cortical dysplasia II A and B, and tubers of tuberous sclerosis, respectively. Also included: dysplasia n.o.s. Not included: focal cortical dysplasia I, which is not diagnosed by neuropathologists in Bonn (Bien et al., 2013). Also not included: temporopolar gray-white abnormalities associated with Ammon's horn sclerosis (Schijns et al., 2011).

  8. g

    Including porencephaly due to perinatal infarction.

Ammon's horn sclerosis, amygdalar sclerosis90232.6540.70
Benign long-term epilepsy-associated tumors (LEATs) with mixed cystic and solid componentsa2478.9411.15
LEATs with predominant solid componentb, malformation tumors, and other low grade gliomas1384.996.23
Oligodendroglioma, oligoastrocytoma411.481.85
Gliomas WHO grade III and IVc371.341.67
Epidermoid cyst80.290.36
Encephalitis (autoimmune, including limbic encephalitis, Rasmussen's)301.091.35
DVA, AVM, vascular n.o.s.d341.231.53
Pial angiomatosise130.470.59
Focal cortical dysplasia including tubers of tuberous sclerosisf29510.6813.31
Nodular heterotopia, subcortical band heterotopia, polymicrogyria, complex brain malformations including lissencephaly, pachygyria, agenesis of corpus callosum, craniocephalic malformations, hemiatrophy, lobar dysgenesis441.591.99
Hypothalamic hamartoma120.430.54
Scars: posttraumatic, postischemicg, posthemorrhagic, postinfectious/abscess, ulegyria, postsurgical n.o.s., glial scar n.o.s.29110.5313.13
No lesion detected54719.8 

MRI-sequence characteristics of epileptogenic lesions

Table 2 documents the suitability of MRI sequences for epileptogenic lesion entities. No single MRI sequence shows high suitability for all entities, but each sequence is highly suitable (++) for at least one of the 16 lesion categories. FLAIR has ++-suitability for more lesion entities than any other sequence (9/16; 56.25%), followed by T2/STIR (7/16; 43.75%), CM-enh. T1 (3/16; 18.75%), T1, IR, hemo/calc (each 2/16; 12.5%), and DWI (1/16; 6.25%).

Table 2. List of epileptogenic entities and the suitability of MRI sequences for their detection
3D-T1T2/STIRFLAIRhemo/calcCM-enh. T1IRDWI
  1. LEAT, long-term epilepsy associated tumors; WHO, World Health Organization; DVA, developmental venous anomaly; AVM, arteriovenous malformation; n.o.s., not otherwise specified.

  2. “++” = sequence allows lesion detection even when the lesion is subtle (high suitability); “+” = larger lesion may be detected, smaller may be missed (moderate suitability). “∘” = even larger lesions can be missed (poor suitability). Suitability rating in brackets: signal characteristics apply to a subset of the respective lesion entities (e.g., meningiomas or oligodendrogliomas with and without calcification; scars with and without hemosiderin deposits).

Ammon's horn sclerosis, amygdalar sclerosis++++
LEATS with mixed cystic and solid components
Cystic component+++++
Solid component+++∘ (+)
LEATs with predominant solid component, malformation tumors, and other low grade gliomas+++
Oligodendroglioma, oligoastrocytoma+++++ (++)∘ (+)
Gliomas WHO grade III and IV++++∘ (+)++++
Meningioma+++∘ (++)++
Epidermoid cyst++++++++
Encephalitis (autoimmune, including limbic encephalitis, Rasmussen's)++++
DVA, AVM+++++++
Pial angiomatosis++++++
Focal cortical dysplasia including tubers of tuberous sclerosis+++++
Nodular heterotopia, subcortical band heterotopia, polymicrogyria, complex brain malformations including lissencephaly, pachygyria, agenesis of corpus callosum, craniocephalic malformations, hemiatrophy, lobar dysgenesis+++++
Hypothalamic hamartoma+++++
Scars: posttraumatic, postischemic, posthemorrhagic, postinfectious/abscess, ulegyria, postsurgical n.o.s., glial scar n.o.s.+++++∘ (++)+

Taking the prevalence of epileptogenic lesion in the Bonn series into account, the superiority of FLAIR is confirmed. It covers 84.8% of the documented lesions with ++-suitability (under the assumption that an estimated 50% in the group of long-term epilepsy-associated tumors (LEATs) with cystic and solid components show predominantly solid, and the other half a predominantly cystic components—a review of the MRIs of all 247 patients was not possible). Second most lesions (62.51%) were identified at ++-suitability level by T2/STIR, follow by hemo/calc (5.78%), CM-enh. T1 (4.15%), T1 and IR (2.98%) each, and finally DWI (0.36%).

Required slice thickness

We identified epileptogenic lesions with therapeutic implications as small as 5–7 mm in diameter. This accounts for example for cavernomas, focal cortical dysplasia type IIB (both resectable with a good chance for seizure freedom) and nodular periventricular heterotopias (suited for high frequency thermocoagulation, Schmitt et al., 2011; Fig. 1A). Usually, a lesion should measure three voxels in one direction to be detectable by MRI. Only in cases of strong contrast between the lesion and the surrounding tissue might a lesion of only two voxels be found. Figure 1B illustrates that despite adequate sequence selection a slice thickness of 4 mm can be too large to identify a small singular periventricular heterotopic nodule.

Figure 1.

(A) Small but therapy relevant epileptogenic lesions. Left: a singular subependymal nodule measuring 0.6 × 1.1 cm. (B) A cavernoma with hemosiderin surrounding measuring 0.5 × 0.7 mm. (C) A histologically proven focal cortical dysplasia type IIB measuring 0.5 × 0.5 mm. All lesions have been overseen on earlier imaging occasions. (D) A singular periventricular nodule is recognizable on the axial 1 × 1 × 1 mm T1 (left, magnification) and also can be recognized on two slices of a coronal 2 mm T2 sequence (a), but not on any slice of a coronal 4 mm STIR sequence (b). A coronally reformatted T1 with 1 mm slice thickness shows the lesion even on five subsequent slices (not shown).

Recommended numbers of slice orientation

We found examples of small lesions in which the pathology is difficult to discriminate from oddly cut physiologic structures when displayed in only one cut plane. The same accounts for the distinction of partial volume effects from pathologic or physiologic structures. Two examples are given in Figure 2. Therefore, the examination should comprise two angulations of each sequence unless a suspect structure can be validated in a second orientation of an equally suited sequence.

Figure 2.

(A) a singular subependymal nodule may be mistaken for a physiologic gyral structure on the axial orientation (A1) but is clearly recognizable when additional coronal (A2) or sagittal (A3) cut plane orientations are available. (B) Larger right than left temporomesial structures on a coronal FLAIR image mimic a swelling of the right hippocampus (full arrow, B1). However, linkage of the coronal with the subsequently performed axial orientation indicates that the coronal slice is tiled and that on the right side the uncus is cut instead of the hippocampal body (B2). Adjusting the tilt according to the axial FLAIR, which demonstrates the hippocampus along its longitudinal axis, indicates that both hippocampi are normal (outlined arrows, B3 and B4).

Recommended slice angulation

On axial FLAIR or T2/STIR scans lesions located outside the temporo-mesial structures where likewise to identify in the hc and the ac-pc angulation. Yet, temporomesial lesions such as hippocampal sclerosis and amygdala swelling (a common sign of limbic encephalitis) were clearly better recognizable on hc-angulated axial scans. In addition, hc-angulated axial scans showing the hippocampus in its longitudinal axis are most suitable to confirm the correct yaw of coronal T2 and FLAIR scans. Figure 2B demonstrates that an odd yaw of coronal scans can lead to misjudgment of amygdala and hippocampal volumes. In summary, hc-angulation of axial T2/STIR and FLAIR scans is recommendable.

In hemo/calc sequences more brain voxels are lost due to susceptiblility artifacts in the ac-pc than in the hc angulation. This indicates that the hc-angulation should be preferentially applied. However, the direct intrasubject comparison of coregistered ac-pc and hc-angulated axial SWI scans indicates more prominent susceptibility artifacts in frontoorbital cortex in the hc-angulation. The clinical relevance of frontoorbital signal drop outs in comparison to signal drop-outs in other localizations is not yet clear. Until studies on this issue exist, we follow the voxel-based statistical evidence and recommend hc-angulated axial hemo/calc scans.

Sensitive but economic battery of sequences

Covering all lesion entities with at least one highly suitable sequence does not require applying all possible sequences (Table 3). In the stepwise sensitivity analysis we found 84.8% of lesions detected with ++-suitability by FLAIR (rank 1). T2/STIR recognizes 6.84% of the remaining patients with ++-suitability (rank 2), thereafter hemo/calc 5.77% (rank 3), T1 and IR likewise 2.98% (rank 4 and 5), and CM-enh. T1 0.59% (rank 6) of the lesions with high suitability, which completes the detection of all lesions.

Table 3. Rank analysis of MRI sequences suited for the detection of epileptogenic lesions
 FLAIRT2/STIRHemo/calcT1/IRCM-enh. T1
  1. Cumulative percentage of lesion entities detected after the first to fifth ranked sequence with regard to their suitability for epileptogenic lesions. For the pure detection of epileptogenic lesions (this study aims at maximizing the sensitivity of MRI for epileptogenic lesions, not the specificity), no further sequences are necessary.

Rank 184.8%    
Rank 2 91.64%   
Rank 3  97.41%  
Rank 4   99.4% 
Rank 5    100%

Proposed epilepsy specific MRI protocol: the “essential six”

The final recommendation for an epilepsy-specific protocol takes several aspects into account. First, the suitability of sequences for epileptogenic lesions: FLAIR, T2/STIR, and hemo/calc together allow for 97.41% of all lesions to be detected. As a fourth sequence an isotropic 3D-T1 must complete the battery of sequences. This is for several reasons: T1-weighted images are suited for migration disorders such as heterotopias, which are important to recognize because they often preclude successful surgery. In contrast to IR, 3D-T1 is a common basis for co-registration of sequential MRI examinations, for MRI postprocessing (for example morphometric MRI analysis for the detection of cortical dysplasias, Huppertz et al., 2005), and for co-registration of MRI with other imaging modalities.

After FLAIR, T2/STIR, hemo/calc, and T1, 99.4% of all lesions and all lesion types are detected except for subtle pial angiomatosis. For the sake of an economic protocol we do not include CM-enh. T1 in our proposal (see 'Discussion').

To capture even small lesions, slice thickness should be minimized to the lowest acceptable value with regard to signal-to-noise ratio. For 1.5 Tesla scanners, this is 1 mm for 3D-T1 and 3 mm for FLAIR, T2/STIR, and hemo/calc sequences. Three Tesla scanners can perform even thinner slices with good signal-to-noise ratio.

To allow distinguishing of small lesions from partial volume effects within a suited sequence, T2 and FLAIR must be acquired in two cut plane orientations. 3D-T1 can be reangulated by most digital picture archiving and communication systems (PACS) and does not need to be acquired in different orientations. For the hemo/calc sequence, axial angulation is sufficient, since questionable findings (mostly cavernomas) can be validated on coronal T1, T2, or FLAIR.

All T2 and FLAIR (coronal and axial) should be angulated perpendicular to the hippocampal axis.

In summary, six specific scans are necessary for an epilepsy-specific outpatient MRI (“essential six”; see Table 4).

Table 4. Epilepsy outpatient specific MRI protocol (essential six)
SequenceSlice thickness (no gap)Cut-plane orientationCut-plane angulation
  1. Economic MRI protocol optimized for the detection of epileptogenic lesions in early outpatient MRI performed in 1.5 Tesla MRI scanners. The T1 sequence should be acquired as 3D data set with isotropic voxel size of 1 × 1 × 1 mm, since this data set offers the best spatial resolution for an acceptable scanning time and can be used to co-register later examinations with the given data set and to perform morphometric analysis. T2, FLAIR, and hemo/calc sequences should be performed at ≤3 mm slice thickness without gap. The field of view of all sequences must cover the whole brain. If applied to 3 Tesla scanners, slice thickness can be further decreased

  2. a1mm isotropic 3D-FLAIR. bThe ideal angulation of axial hemo-calc sequences is subject to further investigation.

  3. Within the sequences different technologies are available, for example 3D T1 magnetization prepared rapid acquisition gradient echo (MPRAGE) or generalized autocalibrating partially parallel acquisitions (GRAPPA) (Lindholm et al., 2009), T2* gradient echo (GRE) or susceptibility weighted imaging (SWI) (de Souza et al., 2008), or conventional Cartesian or PROPELLER T2 or FLAIR sequences (Nyberg 2012).

3D-T11 mm isotropic3 dimensionalac-pc
T2/STIR≤3 mmAxialhc
T2/STIR≤3 mmCoronalhc
FLAIR≤3 mmaAxialhc
FLAIR≤3 mmaCoronalhc
Hemo/calc≤3 mmAxialhcb


Early epilepsy outpatient MRI must find the right balance between the technical possibilities that highly sophisticated MRI examinations offer these days, and the economic constraints of outpatient medicine. It is evident that general radiologists cannot replace specialized MRI examinations at epilepsy centers including MRI postprocessing and application of experimental sequences. Yet, a certain degree of MRI quality must be demanded because otherwise the examination target, the identification of treatable causes of epilepsy, is missed. We believe that an epilepsy-specific protocol for early outpatient MRI should exist that meets the requirements of both epileptologists and radiologists.

The approach chosen in this study is to define an epilepsy-specific outpatient MRI protocol from the spectrum of epileptogenic lesions found in a large cohort of patients with epilepsy who presented for presurgical assessment to one of the largest European epilepsy surgery programs (Bien et al., 2013). In contrast to studies reporting only the histopathologic spectrum of operated patients, this cohort also contains patients in which epilepsy surgery was decided against. This enlarges the list of recognized epileptogenic lesions by often nonoperable etiologies such as polymicrogyria, nodular or band heterotopias, or complex malformations.

We found that not all possible MRI sequences have to be carried out to detect epileptogenic lesions. Four sequences are sufficient for virtually all typical epileptogenic lesions: FLAIR (sensitive for Ammon's horn sclerosis, cortical dysplasia, tumors, inflammations, and scars), T2 (sensitive for cystic tissue components and the internal structure of the hippocampus), T1 (displays cortex-isointense structures), and a hemo/calc-sensitive sequence (highlights lesions with calcification or hemosiderin deposits). The value of a contrast medium–enhanced T1 sequence in an early epilepsy outpatient protocol is debatable. Although it has a place in tumor grading and in the differentiation of vascular lesions, we do not include it in our proposal. The main target of early outpatient MRI is detecting epileptogenic lesions with high sensitivity, and for this CM-enh. T1 is not essential. In our series omission of CM-enh. T1 would have led only to the missing of very subtle pial angiomatosis. Yet, in many patients with epilepsy caused by pial angiomatosis, there is already significant cortical scarring that can be easily displayed on FLAIR or, in case of calcifications, on hemo/calc-sensitive sequences so that the loss of information is certainly less than 0.6% which is the prevalence of pial angiomatosis in the Bonn series. If, in some patients, examinations including contrast-enhanced T1 might be necessary to specify results must be an individualized decision by the responsible radiologist beyond the routine protocol.

A particular challenge of early outpatient MRI is that even small lesions (down to a diameter of ≤5 mm) can cause devastating epilepsy. Because they often have excellent postsurgical outcome, their detection is particularly crucial. Any slices thicker than recommended here or the insertion of gaps between slices increases the risk of missing treatable small lesions. Scanners with 0.5 and 1 Tesla magnetic field strength cannot provide the required thin slices with the sufficient signal-to-noise ratio. For this reason, scanners under 1.5 Tesla must be regarded as obsolete for the imaging of patients with epilepsy, unless no appropriate scanner is accessible. On the other hand, 3 Tesla MRI means a clear improvement in epilepsy imaging. In particular very thin-sliced 3D-FLAIR (for example with isotropic 1mm voxel size, Saini et al., 2010; Lummel et al., 2011) performed at 3 Telsa allows the detection of subtle cortical dysplasias. Also because of small lesions, an epilepsy-specific MRI protocol must contain several cut plane orientiations that can prevent misinterpretation of partial volume effects or odd cut physiologic structures.

The proposed combination of 3D-T1 + axial and coronal FLAIR + axial and coronal T2/STIR + one axial SWI (the “essential 6” sequences) is to our understanding an effective strategy for early epilepsy outpatient imaging.

Although the “essential 6” protocol accords with most published recommendations for MRI protocols in demanding several sequences and slice orientations (Commission on Neuroimaging of the International League Against Epilepsy, 1997; Serles et al., 2003; Deblaere & Achten, 2008; Jackson & Badawy, 2011), it differs from former proposals in some critical points.

  • Some sequences included in earlier recommendations are redundant and can be omitted.
  • In contrast to others protocols (for example Urbach, 2005; Serles et al., 2003) we now propose one set of angulations and orientations for all patients irrespective of a clinical focus hypothesis. The reason for this is that in our experience, outside of centers with special expertise in epilepsy, the focus hypothesis is often incorrect and varies in a given patient between examinations. Varying slice angulations and orientations hinder precise follow-up comparisons.
  • In contrast to other proposals we demand a maximum effective slice thickness of 3 mm for T2/STIR, FLAIR and hemo/calc, and 1mm slices for T1. Earlier recommendations on slice thickness were more liberal.

Performing early outpatient MRI according to the “essential 6” protocol allows for later specialist reevaluation, irrespective of the experience of the first radiologist who reads the MRI or the availability of a sound focus hypothesis at the time of examination. This would make the difference to the 2002 Von Oertzen et al. study in which the yield of expertise reading of outpatient MRI was only marginal in comparison to the yield which resulted from the application of epilepsy specific MRI protocols.

The most obvious disadvantage of the “essential 6” protocol is that it requires more time and is thereby more expensive than “standard brain” MRIs, which allow a turnover time of patients in the scanner of around 20 min. The pure scanning time for the “essential 6” protocol is around 35–40 min in 1.5 Tesla scanners. Although technical progress may speed up the examination to some extent (for example performing the 3D-T1 as “generalized autocalibrating partially parallel acquisition” (GRAPPA), Lindholm et al., 2009), significant shortening of the examination is not possible without decreasing the sensitivity for epileptogenic lesions.

The acceptance of any epilepsy-specific MRI protocol which is as complex as the “essential 6” among radiologists will depend on whether health care funders are willing to compensate the additional costs. Epilepsy is a high cost disease for health care systems (Pugliatti et al., 2007; Yoon et al., 2009). The prospect of saving resources by avoiding unnecessary repeat examinations and the prospect of reducing direct and indirect disease costs by timely and successful epilepsy surgery (King et al., 1997; Langfitt et al., 2007) should be arguments for the health care funders to enter into negotiations on appropriate funding of epilepsy-specific MRIs already at early outpatient stages. Irrespectively, earlier recognition of treatable epileptogenic lesions would be of benefit for patients with pharmacoresistant epilepsy who have increased morbidity and mortality risk, and the social and economic burden of epilepsy (Engel et al., 2012).

Limitations of the Study

The documentation of the patients referred to the epilepsy surgery program of the University of Bonn was done prospectively between 1989 and 2009 (see Bien et al., 2013). However, during this time the MRI scans were not likewise systematically archived. Therefore, Table 2 (reporting the MRI characteristics of lesion entities) does not rely on the retrospective evaluation of the MRIs of the 2,740 patients but on multiple studies addressing the appearance of the respective epileptogenic lesions in MRI.

The proposed protocol is not designed to detect the causes of acute symptomatic seizures. However, its application makes sense early on after initiation of epilepsy, since some treatable causes are missed by acute imaging with computed tomography (CT) or less-specific MRI protocols (e.g., limbic encephalitis).

The proposed protocol represents the technical stand of 2013. If future technical developments allow applying thinner slices with good signal-to-noise ratio even at 1.5 T (for example a 3D-FLAIR at 1 × 1 × 1 mm, which is now available from some MRI manufacturers), the protocol should be adjusted toward recommendations of thinner slices. For 3T scanners, already today thinner slices should be demanded. In addition, in case of development of new sequences, the protocol might require updating.


We thank the Förderverein Ruhr-Epileptology for supporting this study (covering travel costs).


None of the authors has any conflict of interest to disclose. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.