Classification of Temporal Lobe Epilepsy with and without Hippocampal Sclerosis Via Two-level Feature Selection

Abstract

Hippocampal sclerosis (HS) is one of the most common histopathological abnormalities encountered in patients with temporal lobe epilepsy (TLE), which often serves as a diagnosis index of TLE. However, some patients with TLE have no pathologic characteristics of HS, which brings challenge to the diagnosis of TLE. Therefore, exploring effective methods to classify TLE patients with and without HS is meaningful to understanding the pathogenesis of TLE. In this paper, we propose a two-level feature selection method for classification. We select the categories of features as the first level and pick out the discriminating dimensions as the second level. Furthermore, we combine six regional brain characteristics as our features, including regional homogeneity (ReHo), amplitude of low-frequency fluctuation (ALFF), regional functional connectivity strength (RFCS) and three graph-based features. Results show that our method yields higher classification performance compared against the classifiers with single feature and without any level feature selection using functional magnetic resonance imaging (fMRI) data. Moreover, the discriminative brain regions selected by our method are consistent with previous studies. Thus, our method can accurately classify TLE patients with and without HS, which is interpretable from the perspective of physiology at the same time.