English
 
Help Privacy Policy Disclaimer
  Advanced SearchBrowse

Item

ITEM ACTIONSEXPORT

Released

Poster

MRI spectrum of unilateral temporal lobe epilepsy: A surface based pattern analysis of mesiotemporal substructures

MPS-Authors
/persons/resource/persons22828

Bernhardt,  Boris C.
Department Social Neuroscience, MPI for Human Cognitive and Brain Sciences, Max Planck Society;

External Resource
No external resources are shared
Fulltext (restricted access)
There are currently no full texts shared for your IP range.
Fulltext (public)
There are no public fulltexts stored in PuRe
Supplementary Material (public)
There is no public supplementary material available
Citation

Bernhardt, B. C., Kim, H., Bernasconi, A., & Bernasconi, N. (2014). MRI spectrum of unilateral temporal lobe epilepsy: A surface based pattern analysis of mesiotemporal substructures. Poster presented at 68th Annual Meeting of the American Epilepsy Society, Seattle, WA, USA.


Cite as: https://hdl.handle.net/11858/00-001M-0000-002B-2782-E
Abstract
Rationale: Temporal lobe epilepsy (TLE) is the most common drug-resistant epilepsy in adults. On MRI, 70% of patients show hippocampal atrophy (HA) lateralized to the side of the seizure focus. HA relates to seizure-freedom after surgery, with a predictive value surpassing most electro-clinical markers. Hippocampal assessment has influenced presurgical decision-making for decades, reflected in a widely accepted dichotomization of TLE patients based on presence/absence of HA. Increasing evidence, however, challenges this classification. Marked structural changes are present in the adjacent amygdala and entorhinal cortex, with significant individual differences in degree and regional distribution of anomalies. Such variability may hamper post-surgical outcome prediction based on hippocampal assessment alone. Here, unsupervised machine-learning was applied to identify patterns of morphological variability in 134 patients with unilateral TLE. To evaluate the diagnostic validity of our classification, we assessed its relationship to clinical parameters and its yield in the automatic prediction of post-surgical outcome. Methods: We studied 134 patients with drug-resistant TLE (63 males; 16-57 years). 92 patients underwent surgery (mean follow-up=7±2 years). We manually segmented the hippocampus, amygdala, and entorhinal cortex on T1-weighted MRI. A surface-based method measured local volume changes in 12 histologically defined parcels per hemisphere (Fig. 1). Unsupervised k-means clustering aggregated individual patients into distinct classes. The structural markup of TLE classes was compared to a control group (n=46) using surface-based t-tests, corrected for multiple comparisons. We assessed the ability of mesiotemporal surface data and class information to predict post-surgical outcome in the 92 operated patients using linear discriminant analysis (LDA) with leave-one-out cross-validation. Results: Data-driven clustering segregated patients into 4 similarly prevalent classes (Fig. 2). Among patients diagnosed with HA through conventional volumetry, about half of them presented with strictly ipsilateral mesiotemporal atrophy (TLE-II); the other half showed bilateral, yet asymmetric damage (TLE-I). Similarly, patients with normal hippocampal volumetry were generally divided into those with subtle bilateral symmetric atrophy (TLE-III), and a class with a rather paradoxically increased hippocampal and amygdala volumes (TLE-IV). Surface features and class information predicted seizure-freedom in 95%, outperforming conventional hippocampal volumetry (52%, FDR<0.05) and surface mapping without class information (78%, FDR<0.05). Conclusions: Surface-based MRI profiling provides a novel and nuanced characterization of the TLE imaging spectrum. Anomalies were not restricted to the hippocampus alone in any given class; furthermore, except in one class (TLE-I), structural alterations were bilaterally distributed. The clinical validity of our approach is supported by almost perfect accuracy in predicting outcome, outperforming conventional MRI volumetry and class-blinded surface mapping.