Classification and lateralization of temporal lobe epilepsies with and without hippocampal atrophy based on whole-brain automatic MRI segmentation

Abstract

Brain images contain information suitable for automatically sorting subjects into categories such as healthy controls and patients. We sought to identify morphometric criteria for distinguishing controls (n = 28) from patients with unilateral temporal lobe epilepsy (TLE), 60 with and 20 without hippocampal atrophy (TLE-HA and TLE-N, respectively), and for determining the presumed side of seizure onset. The framework employs multi-atlas segmentation to estimate the volumes of 83 brain structures. A kernel-based separability criterion was then used to identify structures whose volumes discriminate between the groups. Next, we applied support vector machines (SVM) to the selected set for classification on the basis of volumes. We also computed pairwise similarities between all subjects and used spectral analysis to convert these into per-subject features. SVM was again applied to these feature data. After training on a subgroup, all TLE-HA patients were correctly distinguished from controls, achieving an accuracy of 96 ± 2% in both classification schemes. For TLE-N patients, the accuracy was 86 ± 2% based on structural volumes and 91 ± 3% using spectral analysis. Structures discriminating between patients and controls were mainly localized ipsilaterally to the presumed seizure focus. For the TLE-HA group, they were mainly in the temporal lobe; for the TLE-N group they included orbitofrontal regions, as well as the ipsilateral substantia nigra. Correct lateralization of the presumed seizure onset zone was achieved using hippocampi and parahippocampal gyri in all TLE-HA patients using either classification scheme; in the TLE-N patients, lateralization was accurate based on structural volumes in 86 ± 4%, and in 94 ± 4% with the spectral analysis approach. Unilateral TLE has imaging features that can be identified automatically, even when they are invisible to human experts. Such morphometric image features may serve as classification and lateralization criteria. The technique also detects unsuspected distinguishing features like the substantia nigra, warranting further study.

Figure 1. The analysis pipeline of the proposed classification scheme.

MAPER: multi-atlas propagation with enhanced registration; RBF: radial basis function; SVM: support vector machine.

Figure 2. Example segmentation result using MAPER.

Coronal section through the T1-weighted 3D MR image of a subject with left hippocampal sclerosis. The left of the subject is shown on the right of the image. Note the clear difference between the atrophic left and normal sized right hippocampus. Other volumetric differences relevant for automatic classification are invisible on visual inspection.

Figure 3. The flowchart of the experiments on assessing the potential bias resulting from the difference in field strength between atlas images and segmentation targets.

A1, A2 and A3: groups of ten subjects from the 30 atlas datasets scanned at 1.5T. group C: ten randomly selected 3T images from the control set. Middle column top row: A1 datasets were used to anatomically segment A2 images with MAPER, resulting in automatically labeled images (A2_secondary). These secondary atlas datasets were then used to segment the A3 images with MAPER. Middle column bottom row: A1 datasets used to segment group C with MAPER. The resulting ten secondarily labeled group C datasets were then used to anatomically segment the A3 images with MAPER. Last column: three sets of anatomical segmentations for A3 images: two automatically generated either via 1.5T or 3T secondary atlases, and one manual gold standard segmentation.

Figure 4. Hippocampal volumes in patients and controls.

Horizontal lines show the medians, boxes indicate interquartile ranges, whiskers show the minimal and maximal values inside the main data, and lozenges show individual values. Blue, right hippocampi; red, left hippocampi. TLE-HA, TLE with hippocampal atrophy; TLE-N; TLE with normal MRI on visual inspection. Suffixes _L and _R denote left and right sided seizure focus, respectively.

Figure 5. Model response curves for Experiment 1 and 4 for two classification schemes.

The classifier accuracy was presented using 83 ranked structures, for each classification experiment (baseline case).