A Longitudinal Imaging and Clinical Data Workflow Identifies Potential Time-Dependent Risk Factors for...

Subarachnoid hemorrhage (SAH) can lead to recurrent seizures, but the clinical factors that influence which patients will develop epilepsy remain unclear. We built a workflow integrating longitudinal clinical, laboratory, and brain imaging data to identify factors associated with post-SAH epilepsy.

Retrospective review of electronic medical records followed by direct patient contact was used to identify patients with nontraumatic SAH and determine whether they later developed epilepsy. A total of 58 longitudinal imaging, clinical, hematologic, coagulation, metabolic, and cerebrospinal fluid (CSF) variables were analyzed from the first 14 days following SAH. Blood location and ventricular size were quantified from serial computed tomography images using an automated pipeline. A two-step statistical approach was used to identify time-dependent variables associated with post-SAH epilepsy. Data-driven logistic regression modeling was implemented using features extracted from individual time series to identify which variables could be the most important for determining post-SAH epilepsy risk.

Out of 134 patients, 15 (11.2%) developed post-SAH epilepsy. Patients who developed epilepsy had persistently lower Glasgow Coma Scores, increased total and pericortical blood volumes up to day 8, as well as elevated peripheral eosinophil counts and systemic immune-inflammatory index in the second week up to day 13 following SAH. Exploratory logistic regression analysis identified systemic immune-inflammatory index, total blood volume, and CSF white blood cells as the most influential variables.

We demonstrate the ability of a novel workflow combining longitudinal imaging and clinical data to identify potential risk factors for post-SAH epilepsy. While limited to a single-center retrospective cohort, we observed that pericortical blood in the first week and increased inflammation in the second week after SAH were associated with development of epilepsy. Our approach could be expanded to include more patients with SAH at multiple sites, inform future clinical trials to prevent post-SAH epilepsy, and explore other neurologic disorders.