Epilepsy is a common and diverse brain disorder, and the underlying mechanisms of its multiple forms and comorbidities are largely unknown. Recent advances in machine learning methods have enabled us to estimate an individual's "brain-age" based on neuroimaging, and this "neuroimaging-based brain-age estimation" is expected as a novel individual-level biomarker for neuropsychiatric disorders. The current study investigated the brain-age for different categories of epilepsy and the following clinical questions: (1) the effects of psychosis on temporal lobe epilepsy (TLE), (2) clinical differentiation for psychogenic non-epileptic seizures (PNES), and (3) clinical discrimination between juvenile myoclonic epilepsy (JME) and progressive myoclonus epilepsy (PME). First, 1196 T1-weighted MRI scans from healthy controls (HC) were used to build a brain-age prediction model by support vector regression, and this model was applied to calculate the brain-predicted age difference (brain-PAD: predicted age-chronological age) in the HC and 318 patients with epilepsy or PNES. Consequently, almost all categories of epilepsy exhibited a significant increase in brain-PAD by over 4 years. TLE with hippocampal sclerosis had a significantly higher brain-PAD than several other categories. In addition, the mean brain-PAD in TLE with inter-ictal psychosis was 10.9 years, whereas that in TLE without psychosis was 5.3 years. The effects of psychosis on the increase in brain-PAD in TLE were significant. PNES demonstrated a comparable mean brain-PAD with that of patients with epilepsy. PME had a higher brain-PAD than JME. This study supports a different brain aging process in epilepsy from healthy subjects, which may be associated with hippocampal sclerosis or inter-ictal psychosis. Furthermore, neuroimaging-based brain-age estimation may provide novel insight into the diverse symptoms of epilepsy. This article also discusses other potential imaging biomarkers for epilepsy.
 Author's abstract