Abstracts

Rationale: To eliminate user-dependencies in applying computer post-processing techniques to structural and functional imagingComputer post-processing facilitates the identification of important regions for epilepsy surgery evaluations.Freely-available utilities have a significant learning curve.It is feasible to automatically combine and leverage the tools in: - BrainSuite - BioImageSuite Web - SPM - FSL, et al.Custom MATLAB code adds further automation/analysis.No user-supervision is necessary. Methods: A MATLAB 'wrapper' controls a postprocessing pipeline: - T1 reference straightening, intensity-normalization, skull stripping (BioImage Suite Web, BrainSuite) - Gray matter segmentation (SPM, FSL, BrainSuite) - Coregistration of PET, DIR, FLAIR, SPECT, or CT with electrodes (BioImage Suite Web) - Well-defined 'Parcels' of specific cortical regions (BrainSuite Surface Volume Registration)MATLAB code analyses image: - Image cortex (gray matter segmentation) for each parcel with intensity, volume and other measures (Figure 2) - Calculation of 'asymmetry index' to identify hypo/hyperintense cortical parcelsAutomatic Ictal-Interictal SPECT Analysis with SPM (ISAS, BioImage Suite Web) (Figure 3)Automatic electrode registration and extraction (BioImage Suite Web) (Figure 3) Results: The entire post-processing and analysis can be performed without user supervision.The process is demonstrated here in a case with a Type 2A focal cortical dysplasia (FCD) not appreciated by conventional methods.9-year-old male with nightly hypermotor seizures and rare convulsions, but normal interictal EEG and nonlocalizing and nonlateralizing ictal EEG. Neuroimaging was interpreted as normal MRI, normal PET and left frontal ictal SPECT. This example retrospectively processed the MRI and PET (8 minutes), ictal and interictal SPECT (2 minutes) and intracranial electrodes (1 minute). Parcellation and MATLAB analysis of PET and FLAIR required another 90 minutes, but correctly identified the small depth of sulcus FCD in all modalities and confirmed the intracranial EEG onset in the depth contacts within the lesion. Conclusions: Computer postprocessing can be available to all clinicians involved in epilepsy surgery evaluations.We demonstrate a method to automate this process using freely available and custom tools.It will be important to validate the method's utility.A web-based application is in development.The method may identify important regions of interest even when conventional interpretation fails to do so, as in this case. Funding: No funding