# KEPPRA: Key Epilepsy Prognostic Parameters with Radiomics in Acute Subdural Hematoma Before Craniotomy

## TL;DR

This study uses radiomic features from CT scans to predict epilepsy risk in patients with acute subdural hematoma before craniotomy.

## Contribution

The study introduces radiomic parameters, particularly elongation, as a novel predictor of epilepsy risk in acute subdural hematoma patients.

## Key findings

- Elongation ≥ 1.45 increases seizure risk 7.78-fold in aSDH patients undergoing craniotomy.
- Radiomic features like elongation showed excellent discriminatory ability (AUC = 0.82) for predicting seizures.
- Univariate analysis found significant associations between radiomic features and clinical factors with epilepsy risk.

## Abstract

Background: Acute subdural hematoma (aSDH) is associated with a high risk of epilepsy, a complication linked to poor outcomes. Craniotomy is a known risk factor, with an epilepsy incidence of approximately 25%. This study evaluated radiomic features from preoperative CT scans to predict epilepsy risk in aSDH patients undergoing craniotomy. Methods: A retrospective analysis of 178 adult aSDH patients treated between 2016 and 2022 identified 64 patients meeting inclusion criteria. Radiomic features (e.g., Feret diameter, elongation, flatness, surface area, and volume) from preoperative CT scans within 24 h of surgery were analyzed alongside clinical factors, including cardiac comorbidities, pupillary response, SOFA score, age, and anticoagulation status. Results: Of the 64 patients, 18 (28%) developed generalized seizures. Univariate analysis showed significant associations with Feret diameter (p = 0.045), elongation (p = 0.005), cardiac comorbidities (p = 0.017), and SOFA score (p = 0.036). ROC analysis showed excellent discriminatory ability for elongation (AUC = 0.82). Multivariate analysis identified elongation as an independent predictor (p = 0.003); elongation ≥ 1.45 increased seizure risk 7.78-fold (OR = 7.778; 95% CI = 1.969–30.723). Conclusions: Radiomic features, particularly elongation, may help predict epilepsy risk in aSDH patients undergoing craniotomy. Prospective validation is needed.