Is streamlined evaluation of children for epilepsy surgery possible?


  • Shekhar G. Patil,

    1. The National Centre for Young People with Epilepsy, Lingfield, United Kingdom
    2. Neurosciences Unit, UCL—Institute of Child Health, London, United Kingdom
    3. Epilepsy Unit
    Search for more papers by this author
  • J. Helen Cross,

    1. Neurosciences Unit, UCL—Institute of Child Health, London, United Kingdom
    2. Epilepsy Unit
    Search for more papers by this author
  • W. Kling Chong,

    1. Department of Radiology, and
    Search for more papers by this author
  • Stewart G. Boyd,

    1. Neurosciences Unit, UCL—Institute of Child Health, London, United Kingdom
    2. Epilepsy Unit
    Search for more papers by this author
  • William J. Harkness,

    1. Neurosciences Unit, UCL—Institute of Child Health, London, United Kingdom
    2. Department of Neurosurgery, Great Ormond Street Hospital for Children NHS Trust, London, United Kingdom
    Search for more papers by this author
  • Brian G. R. Neville,

    1. The National Centre for Young People with Epilepsy, Lingfield, United Kingdom
    2. Neurosciences Unit, UCL—Institute of Child Health, London, United Kingdom
    3. Epilepsy Unit
    Search for more papers by this author
  • Rod C. Scott

    1. The National Centre for Young People with Epilepsy, Lingfield, United Kingdom
    2. Neurosciences Unit, UCL—Institute of Child Health, London, United Kingdom
    3. Epilepsy Unit
    4. Radiology and Physics Unit, UCL—Institute of Child Health, London, United Kingdom
    Search for more papers by this author

Address correspondence to Dr. Rod C. Scott, Senior Lecturer in Paediatric Neurosciences and Honorary Consultant Paediatric Neurologist, Neurosciences Unit, UCL—Institute of Child Health, The Wolfson Centre, Mecklenburgh Square, London WC1N 2AP, United Kingdom. E-mail:


Background: The presurgical evaluation of children with intractable epilepsy includes evaluation by an experienced clinician, MRI, video EEG, and functional imaging techniques to localize seizure onset. However, the contributions of each investigation to surgical decision making has not been systematically assessed.

Method: Data used for decision on eligibility for surgery on 353 children was discussed at a presurgical multidisciplinary meeting and systematically recorded. The relationships between MRI, EEG, SPECT findings, and the probability of being offered epilepsy surgery were investigated retrospectively using a quick unbiased statistical tree (QUEST).

Results: Sixteen children were offered nonresective surgery. Of the remaining, 236 (70%) were offered resective surgery. The proportion of children with a localized lesion on MRI offered resective surgery was 92%[95% CI: 88 to 95%], and EEG telemetry did not modify decision making in this group (p < 0.001). In children with bilateral MRI changes or normal scan the probability of being offered resective surgery was 78% in those with localized ictal onset on EEG compared to 9% with nonlocalized EEG (p < 0.001). SPECT did not appear to systematically influence decision making in any group.

Conclusion: Children with medically intractable epilepsy and localized lesions on MRI may not necessarily need ictal EEG recordings or SPECT prior to offering resective surgery. More targeted use of EEG telemetry could allow more children with less obvious surgical targets to be investigated without increasing resources.

The ILAE subcommission for Pediatric Epilepsy Surgery (Cross et al., 2006) has recently published guidelines for the presurgical evaluation of children with epilepsy. The guidelines suggest that in addition to evaluation by an experienced clinician, EEG (including sleep recording) and magnetic resonance imaging (MRI) are mandatory investigations, and video-EEG telemetry is strongly recommended in all children. Other investigations that may be carried out for seizure localization include ictal and interictal singl photon emission computed tomography (SPECT), positron emission tomography (PET) and magnetoencephalography (MEG). Neuropsychological and neuropsychiatric evaluations are also recommended. Neuropsychological assessment as part of presurgical assessment is essential because it characterizes cognitive and social impairments, which allows assessment of the potential impact of surgery on these domains. However, the relative contributions of the presurgical investigations in the decision on whether to offer epilepsy surgery has not been systematically assessed (Uijl et al., 2005; Berg et al., 2003). Thus, epilepsy surgery in childhood is resource intensive and limited resources may mean that few, if any, countries are able to meet demand.

It is perhaps salutary to look back to the time when surgery for epilepsy began in which seizure semiology, EEG and low-resolution radiological evidence (skull x-rays and sometimes air studies) allowed the identification of mesial temporal sclerosis (MTS) in adults and some children (Falconer et al., 1955). The early history is therefore one of great reliance on the clinical and neurophysiological data given the poor resolution of imaging studies. However with the establishment of sophisticated brain imaging, it is appropriate to reevaluate the relative merits of practices of presurgical evaluation.

The purpose of the current study is to evaluate the basis of making epilepsy surgery decisions in order to assess whether it may be possible to streamline evaluation of children who may benefit from epilepsy surgery. If it were possible to be more selective with regard to investigations then more children may have access to epilepsy surgery without the necessity for large increases in resources.

This retrospective study will allow us to plan a prospective evaluation of the value of preoperative investigations.


The epilepsy surgery program at Great Ormond Street Hospital for Children NHS trust evaluates children mainly from the UK for consideration of surgery for epilepsy. Investigations include clinical review, EEG-video telemetry, MRI optimized for evaluation of children with epilepsy, ictal and interictal SPECT in a selected population, neuropsychometry, and assessment by a neuropsychiatrist. All results are brought to a multidisciplinary meeting to consider whether surgery can be offered. Data presented on individual cases follows a standardized format: clinical summary, EEG telemetry, MRI, nuclear medicine imaging, psychometry assessment and the psychiatric evaluation, discussion, and a report produced. For the purpose of this study, all reports of multidisciplinary meetings were reviewed of children discussed between 1999 and 2005.

Video EEG is recorded with a Grass-Telefactor (Astromed, West Warwick, RI, U.S.A.) system using electrodes placed according to the 10-10 system (Chatrian et al., 1985). MRI is performed on a Siemens 1.5 Tesla system (Siemens, Erlangen, Germany). An epilepsy protocol (Scott et al., 2003) that includes a T1-weighted, 3-dimensional dataset reformatted in the axial and coronal plane at 90° to the long axis of hippocampus, T2-weighted axial and coronal imaging and hippocampal T2 relaxometry was performed. Ictal SPECT (Hartley et al., 2002) imaging was performed during a telemetry admission using the radioisotope 99mTc-Ethyl cysteine dimer that was injected as soon as possible after seizure onset. An interictal SPECT was subsequently carried out. The consultant takes the decision regarding the need for SPECT imaging at the time of initial outpatient assessment and therefore not all children that enter our program undergo this investigation.

For the purposes of the current study the following definitions have been applied.

  • 1) Unilateral structural abnormality—structural lesion affecting a localized or a widespread area of one hemisphere only
  • 2) Bilateral imaging abnormality—structural lesion in both hemispheres
  • 3) Localized—ictal EEG or SPECT with a clear onset in a sub lobar region
  • 4) Lateralized—ictal EEG with an onset in variable or inconsistent regions in one or more lobes within the same hemisphere. SPECT defined as a focus in one or more lobes in same hemisphere.
  • 5) Nonlateralized—no ictal EEG or those with bilateral discharges at the ictal onset.

Statistical analysis

The relationships between MRI, EEG, and SPECT findings, and the probability of being offered epilepsy surgery was investigated using a decision tree. Interictal EEG was not considered in the statistical tree. We used QUEST—Quick Unbiased Efficient Statistical Tree—an algorithm yielding a binary decision tree in SPSS version 13.0 (Chicago, IL, U.S.A.). The model splits data in a binary fashion into smaller subsets with increasing homogeneity with regards to the decision on whether to offer surgery. Our tree was built using a chi-square algorithm and pruned to the minimum tree following which there was no reduction in standard error for prediction of whether surgery will be offered or not. V-fold cross-validation methods were used. We chose to use a QUEST classification tree because of the hierarchical nature of the output, the lack of bias in variable selection for splits and the ability to prune the tree.


The reports of all 353 children evaluated for resective surgery during the 6-year period from 1999 to 2005 were assessed. Of these, 16 were offered palliative surgery and 238 children were offered a procedure, either surgery (both resective and palliative) or invasive monitoring and the remaining 99 were considered ineligible for any form of procedure. The types of resective surgery and the proportions in which they were carried out are shown in Fig. 1. We will concentrate on two groups; those offered a nonpalliative surgical procedure and those considered ineligible for surgery as MRI, EEG, and SPECT/PET are primarily used for localization of a seizure focus and therefore of less relevance in children offered corpus callosotomy or multiple subpial transection (unless they have contributed to the decision not to offer resection).

Figure 1.

Spectrum of resective procedures.

Magnetic resonance imaging

The MRI data of the 353 discussed children is shown in Table 1. Of the 238 children offered a procedure, 215 (90.3%; 95% CI: 86 to 93%) had unilateral and 20 children (8.4%; 95% CI: 5.5 to 12%) had bilateral imaging abnormalities and three had normal imaging (1.2%; 95% CI: 0.4 to 3.6%). The types of abnormalities identified in children with unilateral imaging included dysembryoblastic neuroepithelial tumor (DNET), ganglioglioma, focal cortical dysplasia, hemimegalencephaly, polymicrogyria, MTS, vascular pathology, progressive atrophy of Rasmussen's type, hypothalamic hamartomas, and dual pathologies.

Table 1.  MR imaging in the entire group of 353 children
MRIIneligible for any surgical procedureEligible for a surgical procedurePalliative surgery
Unilateral abnormality18215 0
Bilateral abnormality24 20 0
Normal imaging57  316

Children with bilateral imaging abnormalities who were offered surgery had asymmetric imaging with one cerebral hemisphere more affected than the other and included children with tuberous sclerosis, hypoxic ischemic injury and/or vascular injury.

Of the 99 children not offered a surgical procedure 57 (57.5%; 95% CI: 47.7 to 67%) had normal imaging, 18 (18.1%; 95% CI: 11.8 to 27%) had unilateral imaging abnormality and 24 children (24%; 95% CI: 16.8 to 33%) had bilateral abnormalities.


In the group offered a surgical procedure, ictal events were captured in 228/238 children, 122 (51%; 95% CI: 45 to 57%) had a localized EEG and 59 (24.7%; 95% CI: 19 to 30%) had a lateralized EEG concordant with the hemispheric abnormality or severely affected hemisphere in children with bilateral imaging abnormalities. Three children with unilateral MRI abnormalities had EEG lateralization to the opposite (normal) hemisphere and were offered surgery. The EEG did not lateralize in the remaining 47 children (19.7%; 95% CI: 15 to 25%).

In the children not offered any surgical procedure, a localized EEG was seen in seven children (7%; 95% CI: 3.4 to 13.8%), and a lateralized EEG in 32 children (32%; 95% CI: 24 to 42%).


Ictal SPECT imaging was attempted in 225 children and was successful in 147 (65%; 95% CI: 58 to 71%). A localized uptake of the tracer was seen in 71 (48%; 95%CI: 40 to 56%), and a lateralized uptake in 25 (17%; 95%CI: 12 to 24%) of the scans. Of the 82 successful scans in the group offered surgical procedure, a localized abnormality was seen in 50 (61%; 95% CI: 50 to 71%), and a lateralized abnormality in 12 (15%; 95% CI: 9 to 24%) children respectively.

In the group not offered surgery, of the 65 successful scans, a localized focus was seen in 21 (33%; 95% CI: 23 to 45%), and a lateralized focus in 13 (20%; 95% CI: 12 to 32%) children.

The above data support the view that individually each investigation provides useful information for decision-making, but does not provide information about the independent importance of each test. We therefore proceeded to evaluate this using QUEST. The classification tree is shown in Fig. 2 and the relationships between MRI and video-EEG data in surgical decision making are shown in Fig. 3. Seventy percent of children who entered our surgical program on the basis of a clinical evaluation were offered a surgical procedure following discussion of all data. The first split in the data is on the basis of unilateral structural abnormalities (chi-square = 190.6, df = 2, p < 0.001; See Fig. 2). The proportion of children with unilateral structural brain abnormalities offered surgical procedure is 92.3% (95% CI: 88 to 95%). If the MR imaging is normal or shows bilateral lesions then 78% (95% CI: 71 to 86%) of children will not be offered surgery without further information. In the group with unilateral MR lesions, the model did not produce any further splits on the basis of EEG or SPECT data.

Figure 2.

Classification tree.

Figure 3.

EEG characteristics in 337 children.

Eighteen children in the group with unilateral structural brain abnormalities (See Table 2) were not offered surgery. In 13/18 (72%; 95% CI: 49 to 87%) this was due to involvement of eloquent cortex. Patient 11 had a porencephalic cyst with an ictal EEG onset in opposite hemisphere and porencephaly would not fall into the group of clear focal epileptogenic lesions. However, three children with unilateral structural lesions and EEG documenting ictal onset in opposite hemisphere were offered surgery. There were 2/232 children (0.9%; 95% CI: 0.2 to 3%) with focal MRI abnormalities in whom the EEG and SPECT may have had a role in the decision not to offer surgery. Patient 13 was refused surgery based on the presence of multiple seizure types, a lateralized SPECT involving a wider area than seen on the MRI scan and a nonlocalized EEG. The EEG in patient 14 showed three different areas of onset within the same hemisphere. Of the three children with a clear history of encephalitic illness, none have been offered surgery including patient 14.

Table 2.  Details of the children with unilateral imaging abnormalities who were not offered surgery
NoAgeSexMRIEEGSPECTReason for refusal
  1. R, right; L, left; ND, not done; LOC, localized; LAT, lateralized; NL, nonlateralized; MTS, mesial temporal sclerosis; SWS, Sturge Weber Syndrome.

113y 9mML Middle cerebral artery territory infarctNLNDRisk of functional deficit
216mFSturge Weber syndromeLATNDLow seizure frequency
313yFR Focal cortical dysplasiaNLNDRisk of functional deficit
417y 6mFL perisylvian PolymicrogyriaLOCNDRisk of functional deficit
57y 6mMVascular pathologyNLNDRisk of functional deficit
69y 3mMUnexplained hemisphericLATNDRisk of functional deficit
79y 5mFVascular pathologyNLNDRisk of functional deficit
817y 5mML Focal cortical dysplasiaLATLATRisk of functional deficit
96y 5mFR Focal cortical dysplasiaLATNDRisk of functional deficit
105y 9mMLeft brain atrophyLATNDRisk of functional deficit
118y 11mMLeft Porencephalic cyst—vascular pathologyLAT (Opp)NDDiscordant EEG
1210y 4mMDual pathology—SWS + MTSLOCNLRisk of functional deficit
1310y 11mFDual pathology—MTS + focal cortical dysplasiaLATLATInvestigations suggest widespread damage; not for localized resection.
146yFMature cortical injury—encephalitic illnessNLNDMultifocal onset to her seizures
157y 6mFR Focal cortical dysplasiaNLLOCNeuropsychology & clinical examination—bilateral disease
167y 8mFL PolymicrogyriaLOCLOCRisk of functional deficit
177y 10mFL Focal cortical dysplasiaNLNDRisk of functional deficit
187y 7mMSuspicion of L posterior cortical dysplasiaNLNLRisk of functional deficit & no clear localization

In the group of children with bilateral structural abnormalities or normal imaging, there is a significant role for video-EEG telemetry (chi-square = 45.2, df = 2, p < 0.001. See Fig. 2). Fifteen (79%[95% CI: 57–91%]) children within this group had a localized ictal EEG and were offered surgery, compared to 8 (9.4%[95% CI: 3–15%]) of children with lateralized or nonlateralized ictal EEG. Among the children with bilateral MRI abnormalities and offered a procedure, EEG was localized in 14 (70%; 95% CI: 48 to 86%) and lateralized in 5 (25%; 95% CI: 11 to 47%). Of the 24, not offered any procedure only one child had a localized EEG and this child had a seizure onset in the less obviously affected hemisphere.

Three of the 60 children with normal MRI were offered invasive monitoring and only 1 progressed to surgery. The remaining two showed multiple areas of seizure onset. Of the remaining 57 children, a localized focus on EEG was seen in 3/57 (5%; 95% CI: 1.8 to 14.3%) children, but surgery was refused on grounds of psychometric analysis that suggested bitemporal damage, widespread and nonlocalized abnormality on SPECT and PET scans and risk of inflicting a new functional deficit.


Epilepsy surgery programs are resource intensive. The standard approach is to perform and evaluate a series of diagnostic tests in a stepwise and consecutive manner on a group of children clinically evaluated by an experienced neurologist as to whether surgery may be a way forward. Epilepsy surgery teams clearly develop decision making practices which involve the weight given to specific investigations but have not attempted to measure and discuss them. In the current study the following observations were made:

  • 1) MRI and ictal EEG are individually useful in identifying children with localized onset to their seizures but the investigations should not be considered to be of equal utility. MRI provides the most powerful data for surgical decision making. However there are specific situations when this needs to be supplemented by ictal EEG and/or SPECT particularly with porencephaly, and where the history and imaging suggest an encephalitic illness and the child clearly has multiple seizure types. EEG recordings may be helpful for documentation of the ictal nature of paroxysmal events, although there were no children in our series with exclusively nonepileptic attacks who were refused surgery on those grounds.
  • 2) Ictal EEG and SPECT do not additionally assist in decision making in children with unilateral MRI abnormalities with exceptions as mentioned above.
  • 3) Ictal EEG is useful in making surgical decisions for children with bilateral imaging abnormality.
  • 4) SPECT does not provide routine additional information although its role in special situations may require further evaluation.

Our data suggest that the most important initial investigation is an MRI scan as the presence of a well-defined lesion involving noneloquent cortex such as DNET, or MTS (Cendes et al., 2000; Cukiert et al., 2002). Rasmussen's encephalitis or hemimegalencephaly in a child with a clear history of stereotyped seizures may mean that no further investigation for localization of the seizure focus is required. This group of children accounts for the large majority of children who are offered epilepsy surgery. The spectrum of operations carried out is shown in Fig. 1 which is similar to those published by North American groups (Wyllie et al., 1998; Sinclair et al., 2003).

Given that our outcome data (Doring et al., 1999; Cross, 2002; Devlin et al., 2003; Mclellan et al., 2005) are similar to those reported by other units, and that there is excellent concordance (Haut et al., 2002) across epilepsy surgery programs for the decision to offer epilepsy surgery, a change in approach as suggested above is unlikely to have an impact on outcomes. Such an approach is further supported by the observation that good outcome can be achieved despite discordant MRI and ictal EEG findings in patients with gangliogliomas, DNET and congenital vascular lesions (Morris et al., 1998; Doring et al., 1999; Carreno et al., 2002; Labate et al., 2004). A good outcome can be seen in the presence of generalized ictal EEG recordings in focal lesions and early brain lesions (Gupta et al., 2007; Wyllie et al., 2007). Another study (Holmes et al., 1996) suggests that a proportion of candidates with temporal lobe epilepsy may be offered surgery on consistent interictal EEG recording.

In our current series we have offered surgery to two children with congenital vascular lesions and ictal EEG evidence of apparent seizure onset in “normal” hemisphere but also have been inconsistent as seen with case 11 in Table 2, having not offered surgery on the grounds of a discordant ictal EEG, but in this case took the view that in porencephaly ictal EEG localization was required. Of the two children in the current series who were not offered surgery on the grounds of ictal EEG data, one could have been identified on the basis of multiple seizure types. The other had an encephalitic illness with cortical injury and although our numbers are small, this group may require additional investigation even if there is a unilateral lesion on MRI.

Ictal EEG recording appears to be most appropriate in those patients with bilateral imaging abnormalities. Although this group constitutes a small proportion of children evaluated for epilepsy surgery, a reduction in the number of ictal EEG evaluations in children with focal abnormalities on MRI could mean that there are sufficient resources currently available to evaluate more children with bilateral abnormalities. The other major group of children that could benefit from a freeing up of resources are those with medically intractable focal epilepsy and normal MR imaging. Only one child with normal imaging in the current series was offered resective surgery i.e. (<1%), although a further two had subdural grids placed, but were not found to have a resectable focus. This is comparable to the rate from the Cleveland clinic (Chapman et al., 2005) in which 2.9% of their combined adult and pediatric populations with normal MRI, localized ictal EEG and often concordant nuclear imaging abnormalities were offered surgery. The outcome was favorable in 75%. In another study (Berg et al., 2003), 58/396 cases with normal MRI had surgery. Very few children undergo surgery in the context of normal MR imaging, but it is likely that this population of medically intractable epilepsy in the context of normal MR imaging as seen in adults, (Siegel et al., 2001) may benefit from additional investigations. In our study only 17% of the children discussed at the multidisciplinary meeting had normal MR imaging. However, there are currently no accurate estimates of the prevalence of normal MRI findings in children with intractable epilepsy who are potential surgical candidates to which we can compare our rate (RamchandranNair et al., 2007). This group of children is potentially very important and deserves to be investigated in more detail to identify a seizure focus and the spectrum of clinical phenotypes that may benefit from surgery.

It also has to be stated that we optimized MR imaging to show lesions that could cause epilepsy and how many were delineated after repeated scrutiny cannot be assessed within this study. However, in practical terms this does not matter as children with normal imaging should have ictal EEG recording anyway, and if this serves to help with identification of a structural lesion then this would be considered good use of telemetry.

Although ictal EEG recording may not have a role in evaluation of many children being evaluated for epilepsy surgery, it almost certainly sometimes has relevance in assessing secondary questions. It may be important in evaluating whether apparent multiple seizure types are in fact variations of a single type, whether there are frequent nocturnal events going unreported and therefore giving impetus to carrying out surgery, and in diagnosing nonepileptic attacks. However, ictal EEG recordings may only need to be carried out in those children in whom such questions arise during clinical evaluation. It is also possible that ictal EEG recordings are helpful for prognosis, assessment of timing of operation and predicting the probability of seizure freedom in children with unilateral MRI abnormality, although our own data (Devlin et al., 2003) show that seizure free rates for children undergoing hemispherectomy are comparable to other studies and are not dependent upon ictal EEG findings. The role of ictal EEG for prognosis following epilepsy surgery therefore requires more evaluation. Ictal EEG may be important for deciding the extent of resection in frontal lobe epilepsies (Janszky et al., 2000; Jeha et al., 2007), but again its role in other focal epilepsies remains uncertain. In our series only 27% of children were offered extratemporal resection so those results may be applicable in only one-quarter of our surgical population.

On group analysis in this study, SPECT also appeared to contribute little over and above MRI. It is difficult however, as with ictal EEG recording to determine individual contribution made in specific cases. In addition, the role of SPECT in certain circumstances e.g. in children with no MRI abnormality, requires further evaluation. It is likely that evaluation by an experienced clinician, as well as the gathering of important information provided by psychometry and psychiatry assessments, is essential for appropriate decision making. Psychological assessment as part of presurgical assessment is essential for the characterization of cognitive and social impairments, which allows assessment of the potential impact of surgery on these domains. As can be seen from our data, the information is rarely used to refuse the option of surgery although three children were not offered surgery on the grounds of an assessment that suggested abnormal function in both temporal lobes. Similarly a psychiatric assessment usually guides the surgeon to premorbid disorders and may provide predictive information for psychiatric problems postsurgery. We also use this assessment to clarify the aims and expectations of surgery with the family.

Epilepsy surgery programs are resource intensive enterprises. Our data suggest that ictal EEG and ictal SPECT scans may not be necessary in all the children coming through surgery programs and it may be possible to streamline presurgical evaluation in a way that will allow access of more children to such programs without having to increase available resources. However it is important that other units evaluate their own programs to validate the findings in our study prior to widespread implementation of its findings. In addition we are conducting a prospective assessment of the value of telemetry in our hands. We recognize that there is reassurance for both doctors and parents in having captured a typical seizure but how that helps our decision making process needs to be evaluated. The legal issue would rest on the strength of the evidence for the value of telemetry in specific situations and the current practice of major units.


Dr Shekhar Patil is the recipient of a research fellowship from The National Centre for Young People with Epilepsy, Lingfield, UK. Funding sources played no part in the preparation of this review or in the decision to submit it for publication. Research at UCL—Institute of Child Health and Great Ormond Street Hospital for Children NHS Trust benefits from R&D funding received from the NHS Executive.

We would also like to acknowledge other members of the team; Dr. Steve White, Dr. KB Das, Dr. Sarah Aylett, Dr. Paola Nicolaides, Dr. Dawn Saunders, Dr. Tim Cox, Dr. Isobel Heyman, Sue Harrison who played important role in the discussion at our multidisciplinary meeting.

Conflict of interest: As authors of this work, “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.” The authors have no conflicts of interest to declare.