New MRI-Based Classifier Accurately Lateralizes Hippocampal Pathology in Temporal Lobe Epilepsy

Andrea Bernasconi, MD

Using a newly developed classifier that combined T1-weighted and T2-weighted and fluid-attenuated inversion recovery (FLAIR)/T1 features, investigators were able to accurately determine sides of hippocampal pathology in surgical candidates with temporal lobe epilepsy (TLE).

Published in Neurology, the study featured a training set that included 60 patients with TLE (mean age, 35.6 years; 58% female) with histologically verified hippocampal sclerosis (HS), followed by 2 additional validation cohorts with of patients with similar demographics and electroclinical characteristics to further test accuracy (n = 57; MRI-negative 58%). In the training cohort, the MRI was reported as unremarkable in 42% (n = 25) of the cases, with only 1 additional patient later turning MRI-positive after performing volumetry.

Senior author Andrea Bernasconi, MD, professor of neurology and neurosurgery, McGill University; and co-founder and co-director, Neuroimaging of Epilepsy Laboratory, McConnell Brain Imaging Center, Montreal Neurological Institute, and colleagues documented that converging these diagnostic tests yielded an overall lateralization accuracy of 93% (95% CI, 92-94), regardless of HS visibility.

READ MORE: Newborn Screening, mTOR Pathway May Offer New Hope for Tuberous Sclerosis

In the training cohort, FLAIR/T1 (85% [±1]) and T2 signal (86% [±2]) were superior to that of volumetry (77% [±3]; P <.05). More specifically, the combination of FLAIR/T1 and T2 signal had an accuracy of 84% (±5) in MRI-negative patients with TLE and 100% (±1) in those who were MRI-positive. In both MRI-negative and MRI-positive patients, the classifier was consistently correct across 90 of 100 iterations.

"Conversely, because the presurgical evaluation of drug-resistant TLE is multidisciplinary, an MRI-derived binary lateralization outcome (right vs left) may not be sufficient per se for surgical decision-making,” the study authors wrote. "However, the increased availability of imaging-derived classification algorithms ought to pave the way for systems that integrate diverse sources of evidence, including other imaging modalities such as PET, as well as electroclinical data to increase diagnostic yield and certainty."

The combination of T2 and FLAIR/T1 continued to yield the best overall performance in the validation cohorts, with an over 90% lateralization accuracy. Moreover, the area under the curve (AUC) in validation cohorts 1 (1.00 [±0.00]) and 2 (0.94 [±0.12]) showed the discriminative capability of this combination in MRI-negative patients.

A group analysis of patients included in the study showed that FLAIR/T1 was increased across all subregions (CA1–CA3: t = 3.5 [PFWE <.0001]; CA4–DG: t = 4.3 [PFWE = .004]; subiculum: t = 2.4 [PFWE <.0001]), with additional subtle increases contralaterally. Anomalies in MRI-positive patients had similar distributions across subfields, albeit more severe, according to study authors (volume: t = −3.9 to −4.4 [PFWE <.0001]; T2 signal: t = 3.6-5.4 [PFWE <.0001]; FLAIR/T1: t = 2.7-6.0 [PFWE <.002]).

There was no volumetric alteration in MRI-negative patients; however, they did show subtle increases in the ipsilateral CA4 to DG (t = 2.5; PFWE <.0001) and CA1 through CA3 (t = 2.1; PFWE = .014). These patients also had FLAIR/T1 hyperintensities along all subfields (CA1 through CA3: t = 2.5 [PFWE <.0001]; CA4– DG: t = 2.9 [PFWE = .005]; subiculum: t = 2.1 [PFWE <.0001]).

Bernasconi concluded that the methodology presented “may be expanded to other epilepsy syndromes associated with HS, including cortical developmental malformations.” Bernasconi, along with study coauthor Neda Bernasconi, MD, were presented with the 2021 Clinical Sciences Research Award by the American Epilepsy Society at its 2021 annual meeting, December 3-7, in Chicago, Illinois.² One of the highest research awards, this is given to those whose research contributes importantly to understanding and conquering epilepsy. The 2 co-founded and currenty co-direct the Neuroimaging of Epilepsy Laboratory at McGill University together.

REFERENCES
1. Caldairou B, Foit NA, Mutti C, et al. MRI-based machine learning prediction framework to lateralize hippocampal sclerosis in patients with temporal lobe epilepsy. Neurology. 2021;97(16): e1583-e1593. doi:10.1212/WNL.0000000000012699
2. Neda Bernasconi, MD, PhD, and Andrea Bernasconi, MD, receive the 2021 Clinical Science Research Award. News release. American Epilepsy Society. December 6, 2021. Accessed January 13, 2021. https://www.aesnet.org/about/press-room/neda-bernasconi-md-phd-and-andrea-bernasconi-md-receive-the-2021-clinical-science-research-award