Development of multimodal neuroimaging models for lateralization of temporal lobe epilepsy
M. Nazem-Zadeh, K. Elisevich, H. Soltanian-Zadeh
Apr 1, 2018
Citations
0
Citations
Quality indicators
Journal
2018 3rd Biennial South African Biomedical Engineering Conference (SAIBMEC)
Full text
Semantic Scholar
Key Takeaway Key Takeaway: The developed multimodal neuroimaging model effectively lateralizes temporal lobe epilepsy, potentially eliminating the need for invasive recording in some cases.
Abstract
Response-driven multimodal models were developed to lateralize temporal lobe epilepsy (TLE). Neuroimaging features of 138 retrospective TLE patients with Engel class l surgical outcomes were extracted, including the hippocampal volumes, normalized ictal-interictal SPECT and MR FLAIR intensities, mean diffusivity, and cingulate and forniceal fractional anisotropy (FA). Using logistic function regression, univariate and multivariate models were developed. The model incorporated all multivariate attributes for 138 TLE cases that had at least one imaging attribute imputing the missing attributes with the mean values of the corresponding attributes measured in a control cohort. A probability of detection and false alarm of 0.83 and 0.17, respectively, was attained for all cases. For patients undergoing phase II intracranial monitoring, these metrics were 0.90 and 0.10. The proposed multimodal neuroimaging model for lateralization of TLE is relevant for surgical decision-making and, in some cases, may eliminate the need for invasive recording.