Application of Statistical Parametric Mapping in Children with Medically Intractable Epilepsy

Objectives Very few published studies exist that use SPM to evaluate cerebral FDG scans in paediatric patients with medically intractable epilepsy prior to surgery, largely due to lack of a suitable control dataset. The aim of this study is to implement SPM analyses of the pre-operative cerebral FDG scan in children who subsequently underwent resective epilepsy surgery. The hypothesis is that SPM identification of a hypo-metabolic focus localising to the subsequent surgical resection zone (SRZ) will predict a good response to surgery as measured by improved seizure control.

Methods Using specificity optimised SPM parameters; a single-subject versus group SPM analysis was performed on the pre-operative cerebral FDG scan in 61 children with medically intractable epilepsy that subsequently underwent resective epilepsy surgery. The pre-operative scan of each patient was compared against a high quality control dataset of 28 children with no neurological conditions. SPM results were categorized according to the strength of the result in localising a hypo-metabolic focus to the SRZ. The diagnostic test performance of SPM localising a hypo-metabolic focus to the SRZ in predicting good post-operative seizure control was evaluated (sensitivity (SN), specificity (SP), positive (PLR) and negative likelihood ratios (NLR), positive (PPV) and negative predictive vales (NPV), and area under the curve (AUC) of the receiver operator characteristic (ROC) curve).

Results SPM localizing a hypo-metabolic focus to the SRZ had a very high specificity for predicting good post-operative seizure control. There were persisting false negative results that could be attributable to unidentified electrical seizure activity during the FDG uptake phase. SPM performed best with voxel-wise FWEp correction for multiple comparisons and including all patients the diagnostic test metrics were: SN 38%, SP 93%, PLR 5.4, NLR 0.7, PPV 95%, NPV 31%, AUC of ROC curve 0.638. The diagnostic test performance of SPM was maximal in patients with extra-temporal resections after excluding those with EEG or clinical seizures during FDG uptake.

Conclusions As a stand-alone test, SPM performs moderately well in predicting that a planned surgical resection site will result in improved post-operative seizure control. The poor sensitivity may be due to the large proportion of patients who did not have scalp surface EEG recorded during FDG uptake to allow detection of subclinical electrical seizure activity that would disrupt cerebral FDG uptake. This analysis is the largest reported cohort of dedicated FDG scans in childhood epilepsy and is the only SPM study in children that uses a true paediatric control dataset consisting of children with no neurological abnormalities. The results indicate that SPM analysis of pre-operative FDG PET scans when used as a complementary test with other imaging modalities is useful when localising a hypo-metabolic focus to the planned SRZ, to confirm a high probability of post-operative benefit.