Leveraging pathological markers of lower grade glioma to predict the occurrence of secondary epilepsy, a retrospective study.

Epilepsy is a common manifestation in patients with lower grade glioma (LGG), often presenting as the initial symptom in approximately 70% of cases. This study aimed to identify clinical and pathological markers for epileptic seizures in patients with LGG. Additionally, it sought to develop and validate a machine learning model that enables tailored risk-based anti-seizure treatment. Health records of patients with histologically confirmed LGG from 2019 to 2022 were retrospectively analyzed, incorporating patient demographics, tumor pathology, and epilepsy prevalence data. A random forest (RF) model (named SEEPPR) was constructed based on potential risk factors associated with epilepsy in LGG patients. Performance was evaluated using the area under the receiver operating characteristic (ROC) curve with the SEEPPR model, while the SHapley Additive exPlanation (SHAP) method was employed for elucidating the model's decision process. Additionally, the model has been integrated into a web application to enhance its clinical utility. This study identifies specific clinical and pathological markers as epileptic drivers. Our explainable RF model effectively predicts secondary epilepsy risk in LGG patients, potentially enabling early intervention to prevent epilepsy progression. This study underscores the significance of leveraging machine learning models to enhance epilepsy management in LGG patients.