Using machine learning on MRI data from 296 individuals with temporal lobe epilepsy (TLE) and 91 healthy controls, this study identified four TLE subtypes based on distinct patterns of brain atrophy progression. Two subtypes showed hippocampus-predominant atrophy, one was cortex-predominant, and the fourth exhibited amygdala enlargement without atrophy. These subtypes differed in neuroanatomical signatures, disease progression, epilepsy characteristics, and five-year seizure outcomes after surgery or medication, suggesting a diverse pathobiological basis for TLE with implications for personalized medicine.
