Age-Specific 18F-FDG Image Processing Pipelines and Analysis Are Essential for Individual Mapping of Seizure Foci in Paediatric Patients with Intractable Epilepsy

Fluoro-18-deoxyglucose positron emission tomography (FDG-PET) is an important tool for the pre-surgical assessment of children with drug-resistant epilepsy. Standard assessment is carried out visually and this is often subjective and highly user-dependent. Voxel-wise statistics can be used to remove user-dependent biases by automatically identifying areas of significant hypo/hyper-metabolism, associated to the epileptogenic area. In the clinical settings, this analysis is carried out using commercially available software. These software packages suffer from two main limitations when applied to paediatric PET data: 1) paediatric scans are spatially normalised to an adult standard template and 2) statistical comparisons use an adult control dataset. The aim of this work is to provide a reliable observer-independent pipeline for the analysis of paediatric FDG-PET scans, as part of pre-surgical planning in epilepsy. METHODS: A pseudo-control dataset (n = 19 for 6-9y, n = 93 for 10-20y) was used to create two age-specific FDG-PET paediatric templates in standard paediatric space. The FDG-PET scans of 46 epilepsy patients (n = 16 for 6-9y, n = 30 for 10-17y) were retrospectively collated and analysed using voxel-wise statistics. This was implemented with the standard pipeline available in the commercial software Scenium and an in-house Statistical Parametric Mapping v.8 (SPM8) pipeline (including age-specific paediatric templates and normal database). A kappa test was used to assess the level of agreement between findings of voxel-wise analyses and the clinical diagnosis of each patient. The SPM8 pipeline was further validated using post-surgical seizure-free patients. RESULTS: Improved agreement with the clinical diagnosis was reported using SPM8, in terms of focus localisation, especially for the younger patient group: kScenium=0.489 versus kSPM=0.805. The proposed pipeline also showed a sensitivity of ~70% in both age ranges, for the localisation of hypo-metabolic areas on paediatric FDG-PET scans in post-surgical seizure-free patients. CONCLUSION: We show that by creating age-specific templates and using paediatric control databases, our pipeline provides an accurate and sensitive semi-quantitative method for assessing FDG-PET scans of patients under 18y