Authors :
Presenting Author: Judy Chen, BHSc – McGill University
Jessica Royer, PsyD – PhD Student, Neurology and Neurosciences, McGill University; Sara LaRiviere, PhD – Postdoc, saratheriver@gmail.com, saratheriver@gmail.com; Alex Ngo, BSc – Masters Student, Neurology and Neurosciences, McGIll University; Raul Rodriguez-Cruces, PhD – Postdoc, Neurology and Neurosciences, McGill University; Jordan DeKraker, PhD – Postdoc, Neurology and Neurosciences, McGill University; Andrea Bernasconi, MD – Professor, Neurology and Neurosciences, McGill University; Neda Bernasconi, MD PhD – Professor, Neurology and Neurosciences, McGill University; Dewi Schrader, MD – BC Children's Hospital – University of British Columbia; Boris Bernhardt, PhD – Professor, Neurology and Neurosciences, McGill University
Rationale:
Approximately one-third of children diagnosed with epilepsy are drug-resistant (DRE), with surgical resection the efficient treatment for seizure control. Clinical heterogeneity in pediatric patients, however, can make it difficult to identify a surgical target. The application of post-processing analysis tools that capitalize on multimodal magnetic resonance imaging (MRI) data on adult cohorts has demonstrated promising outcomes in lesion lateralization and localization, alongside potential clinical utility. In this study, we leverage these open-source tools to showcase validity and generalizability on a single-site pediatric focal epilepsy cohort.
Methods:
A cohort of eight healthy controls (5/3 M/F, mean ± standard deviation age=12.5±3.45 years) and six patients with temporal lobe epilepsy (TLE) (2/4 M/F, age=12±3.9 years) were recruited from BC Children’s Hospital. All individuals underwent multimodal 3T MRI imaging including: T1-weighted (T1w), quantitative T1 (qT1), diffusion-weighted, and fluid-attenuated inversion recovery (FLAIR). Imaging data were processed using micapipe (Rodriguez-Cruces et al., 2022) and hippunfold (DeKraker et al., 2022). Results were mapped onto standardized cortical, subcortical, and hippocampal surfaces. Further post-processing using z-brains (https://github.com/MICA-MNI/z-brains) computed individualized asymmetry measures of morphology (cortical thickness, subcortical volume, hippocampal thickness), microstructure (apparent diffusion coefficient [ADC], fractional anisotropy [FA]), and modality-specific signal intensity (qT1, FLAIR). Data in each patient were z-scored against the control cohort.
Results:
We highlight two children diagnosed with TLE different underlying pathophysiological processes (Fig1); PX012 with a left temporal ganglioglioma and PX015 with right-sided mesial temporal sclerosis. Subcortical and hippocampal measures along with corresponding asymmetry analyses best predicted lateralization, however, ADC-specific intensity and asymmetry outperformed other modalities for PX012 (Fig1A) while volume-specific differences and asymmetry were most apparent for PX015 (Fig1B), likely reflecting pathophysiological heterogeneity. Cortical asymmetry maps outperformed regional z-scored maps at identifying suspected lesion location across all modalities. More specifically, T1w and diffusion modalities were most successful in lateralization and lesion mapping across all patients.
Conclusions:
Multimodal imaging with z-brains post-processing analyses provide clinically useful quantitative and visually intuitive representations to aid clinicians in tailoring treatment strategies to individual patients despite heterogeneity in paediatric focal epilepsies. Multi-site, prospective pediatric cohorts and post-surgical data are necessary to further assess tool validation and generalizability.
Funding: FRQS