MRI derived hippocampal asymmetry identifies hippocampal sclerosis in epilepsy surgical specimens

Abstract

Abstract Hippocampal sclerosis is the most common histopathological diagnosis in epilepsy surgical specimens. Although hippocampal sclerosis can be identified in vivo with brain MRI, it cannot be detected by expert qualitative review in as many as 50% of cases. Quantitative features derived from structural brain MRI, such as hippocampal shape, volume, and asymmetry features, show promise in identification of hippocampal sclerosis in vivo but have had limited translation to clinical practice partly due to the need for large normative cohorts for estimation of hippocampal pathology. Using the Imaging Database for Epilepsy And Surgery cohort, which includes brain MRI scans from 442 individuals with drug-resistant focal epilepsy who underwent epilepsy surgery and 100 healthy controls acquired on multiple scanners, we evaluated the diagnostic performance of the hippocampal asymmetry index calculated using only the participant’s hippocampal volumes against a gold standard histopathological diagnosis of hippocampal sclerosis. The hippocampal asymmetry index had high diagnostic performance (area under the curve for left hippocampal sclerosis = 0.95 (95% confidence interval 0.92–0.97); right hippocampal sclerosis = 0.96 (95% confidence interval 0.94–0.98) and was similar regardless of segmentation algorithm, data harmonization, and use of normative data. To facilitate validation, we provide an open-source software tool for the calculation of the hippocampal asymmetry index.