Neuro Informatics Support Analyst Seattle Children's Hospital Seattle, Washington
Rationale: Evaluation of possible seizure onset zones in surgically remediable epilepsy is often a serial process of stacking up critical sets of images from multiple imaging modalities including MRI, PET, fMRI, and CT. Abnormal loci identified on structural MR (e.g., dysplasia) may correlate with hypometabolism on PET, or activation on fMRI. These data are commonly reviewed separately and over time by epileptologists, radiologists, and neurosurgeons, with conference discussions often being the main forum for achieving consensus. For both planning and postoperative visualizations, electrode overlay is often displayed on at most one image type. As a solution to combat this data diffusion, and to facilitate execution of subsequent surgical steps, we developed software with the following features: a) registration of multi-modal imaging in a central sandbox, b) efficient and precise localization of SEEG electrode contacts with automated gray/white matter quantification and an embedded atlas, and c) the capability to generate composite images for subsequent surgical navigation (e.g., critical SEEG contacts overlaid on a T1). We report on the software’s utilization on 17 recent SEEG cases. Methods: We built custom SEEG-focused modules on top of the open source imaging toolbox 3D Slicer. Our software includes an integrated electrode browser, allowing interactive and facilitated identification of SEEG electrode contact locations guided by planned trajectories. Combined with multilayered review of all modes of imaging, and one-click navigation to any contact location, it provides views parallel and perpendicular to the electrode trajectory (Figure 1). The primary strategy developed registers high resolution CTs and contrast MRs to a non-contrast MR, PET imaging to a high-resolution CT via the associated attenuation correction CT, and fMRI data to the associated structural MR image. A fully automated FreeSurfer pipeline was used for subject-specific anatomical and gray/white label map generation. Results: Use of developed tools on 17 SEEG patients demonstrated that image registration was always achievable using the described strategy. Using a pre-SEEG non-contrast T1, subject-specific atlas generation took ~7.5 hours on a standard workstation, during which time all pre-operative imaging can also be co-registered. As this work is complete prior to SEEG implantation, the remaining steps (registration of post-SEEG CT, facilitated identification of contact locations, and generation of anatomical and graphical annotations) can be completed in as little as 1 hr after the completion of the SEEG procedure. Knowledge of the planning trajectories allowed evaluation of the correspondence between planned and placed electrode contact locations (Table 1), confirming that placement accuracy at our institution is comparable to that found in the literature (e.g., Brandmeir 2018). In four cases when patients underwent resection, new overlays including key SEEG contact locations were generated and utilized in image-guided surgical navigation. Conclusions: Co-registration of multimodal imaging with SEEG contacts, combined with easy navigation, allows flexible exploration of relationships between multimodal imaging findings and correlation between imaging and icEEG data. Export of key contacts locations into image volumes facilitates subsequent image-guided surgery. This tool also enables investigation of potential biomarkers in relation to molecular pathology and outcome. The imaging toolbox presented here provides a facilitated workflow for pre- and post-SEEG investigations. Funding: Please list any funding that was received in support of this abstract.: Supported by the Eplilepsy Research Fund through the Center for Integrative Brain Research.