Damage to serotonergic and opioid networks relates to post-stroke epilepsy after thrombectomy

Post-stroke epilepsy (PSE) is a clinically relevant complication after ischemic stroke. While lesion location and established clinical risk factors contribute to PSE risk, the role of lesion-induced disruption of neurotransmitter-specific brain networks remains unclear. We retrospectively analyzed 251 patients with acute large-vessel occlusion ischemic stroke treated with mechanical thrombectomy. Binary lesion masks were embedded into normative neurotransmitter-informed structural connectomes derived from PET-based receptor and transporter density maps, yielding damage scores for 19 neurotransmitter systems and a global measure of structural disconnection. Partial least squares (PLS) regression was used for feature selection, followed by multivariable logistic regression adjusted for age, sex, and SeLECT. As a secondary internal resampling analysis, elastic-net logistic regression with repeated stratified cross-validation was used to assess whether the identified pattern remained informative under regularization. Twenty-six patients (10.4%) developed PSE. PLS identified a neurochemical signature dominated by serotonergic and {micro}-opioid-informed systems. In adjusted models, damage to 5-HT1a, 5-HT2a, and {micro}-opioid networks showed the strongest and most robust associations with PSE, independent of clinical predictors and global structural disconnection. Cholinergic network measures showed weaker and less consistent effects. In repeated internal cross-validation, the same networks were selected more consistently and were associated with higher discrimination than the clinical base model. Lesion-induced disruption of specific neurotransmitter-informed structural networks, most robustly serotonergic and {micro}-opioid systems, is associated with PSE and showed incremental signal in internal cross-validation beyond established clinical risk factors. These findings provide a mechanistically interpretable extension to existing PSE risk models.