فاطمه شاه محمدي مهرجردي

Epilepsy, a chronic neurological disorder characterized by recurrent seizures, affects millions worldwide, with status epilepticus (SE) representing a severe form that requires urgent intervention to prevent long-term brain damage or death. While traditional seizure assessment methods, such as Racine's scale and EEG analysis, provide valuable insights, they lack automation for early SE detection. This study aims to classify seizure stages and detect early SE using a rat model, marking the first known attempt of its kind. Five male Wistar rats underwent a PTZ-induced SE protocol, with intracranial EEG data recorded from four channels located in the CA1 and CA3 regions of the hippocampus in both hemispheres. After preprocessing the data, three types of features were extracted: time-frequency domain features, nonlinear dynamics features, and features related to the connectivity and causality between channels. These features were input into a convolutional neural network (CNN) for classification, and the results were eva‎luated across five different scenarios: (1) a 6-class classification to assess the separability of seizure waveforms at different stages, (2) a 5-class classification to eva‎luate whether the stage intervals were distinguishable, (3) a 3-class classification with non-convulsive, pre-convulsive, and convulsive categories, and (4) a 2-class classification differentiating non-convulsive from convulsive phases. All features performed effectively across these scenarios, achieving 100% accuracy in distinguishing non-convulsive from convulsive stages, with the time-frequency features yielding the best results overall across all classification scenarios.

بحران‌صرع , مراحل‌بحران‌صرع , تجزيه و تحليل زمان-فركانس , ديناميك غيرخطي , ارتباطات مغزي , مقياس اصلاح‌شده ريسين , شبكه‌عصبي كانولوشن

status epilepticus , status epilepticus stages , Time-frequency analysis , non-linear dynamics , Brain connectivity , revised Racine's scale , Convolutional Neural Network

Fateme Shahmohamadi

Dr. Erfanian