Accuracy of High Resolution T1-MRI Tissue Types with EEG-Based Lateralization for Pediatric Patients with Non-Lesional Focal Epilepsy
Abstract number :
2.154
Submission category :
5. Neuro Imaging / 5A. Structural Imaging
Year :
2019
Submission ID :
2421601
Source :
www.aesnet.org
Presentation date :
12/8/2019 4:04:48 PM
Published date :
Nov 25, 2019, 12:14 PM
Authors :
Jayoung Pak, Rutgers University, NJMS; Siddharth Gupta, Rutgers University, NJMS; Oluwafayajimi Salako, Rutgers University, NJMS; Luke Tomycz, Rutgers University, NJMS; Nasrin Ghesani, Mount Sinai Medical Center; Sridhar Kannurpatti, Rutgers University, N
Rationale: Localization of seizure network in pediatric patients with non-lesional focal epilepsy has been challenging. Brain MRI is useful, however atlas transformed structural volumetric analysis of tissue classes such as cerebrospinal fluid (CSF), gray matter (GM) and white matter (WM) are potential biomarkers to lateralize the seizure hemisphere and further localize the seizure networks. In this study, an MRI based blind prediction of epilepsy lateralization using high resolution T1-weighted MRI of 35 pediatric patients with epilepsy were performed against 40 age matched controls. Prediction accuracy was tested against scalp electroencephalography (EEG) based lateralization using an assessment of ictal and interictal epileptiform discharges. Methods: Pediatric patients with a non-lesional brain MRI and focal onset epilepsy were included in the study. MRI images were spatially transformed to the standard Talairach Atlas template after skull stripping using AFNI (Analysis of Functional NeuroImages). Brain images were automatically segmented to distinguish CSF, GM and WM tissue types using the FSL software. Hemispheric lateralization of seizures was predicted from volumes of each segmented tissue type with appropriate atlas based hemispheric masks. MRI T1 image data of 40 control subjects were obtained from the open source Autism Brain Imaging Data Exchange (ABIDE I) study database. Mean ± SD between the normal control subjects in CSF, GM, WM volumes and a composite ratio [(GM+WM)/CSF)] were 3.56±2.85 (cm3), 3.54 ± 2.46 (cm3), 1.94 ± 1.75 (cm3) and 0.3 ± 0.24 respectively. Control subject parameters were normally distributed; hence tissue types of study patients were tested for a significant difference in hemisphere structural parameters from the control population data using a two tailed t-test. Results: In all 35 patients, prediction accuracy of GM and WM was 68% and CSF was 55%. The composite ratio predicted with a 78% accuracy. In a subset of 25 patients, hippocampal volumetry was performed. While hippocampal non-GM/WM tissue led to 67% accurate prediction, hippocampal GM was the highest with 80% accuracy followed by hippocampal composite ratio of 71%. Hippocampal WM had 58% prediction accuracy. A conjunction analysis using the two highest accuracy predictors, hemispheric composite ratio and hippocampal GM, was possible across 21 patients which led to correct prediction of hemispheric seizure lateralization in 18 patients with an accuracy of 86%. Conclusions: Structural volumetric analyses of high resolution T1-MRI can support EEG based hemispheric seizure lateralization of pediatric patients, furthermore the composite ratio of the different tissue types increases prediction accuracy. Funding: No funding
Neuro Imaging