| Abstrakt: |
Early detection and proper treatment of epilepsy seizure is essential and meaningful to those who suffer from this disease. Symptoms of seizures are confusion, abnormal gazing, and obsessive hand movements. Neurological exams, blood testing, cognitive tests, and neuroimaging modalities are some of the techniques used to detect epileptic seizure activity in patients. Among them, neuroimaging techniques have attracted a lot of interest from doctors in the field. The use of computer-aided diagnostic systems (CADS) based on deep learning (DL) and neuroimaging modalities can help quickly and accurately diagnose epileptic seizures. This research introduces a novel seizure detection approach based on a bidirectional long short-term memory (Bi-LSTM) network. As a result, Bi-LSTM preserves the nonstationary nature of electroencephalogram (EEG) data while decreasing processing costs by employing local mean decomposition (LMD) and statistical feature extraction algorithms. Two distinct LSTM networks with opposing propagation directions are combined in the deep architecture. So, the deep model can use information from both before and after the current time of analysis to decide the output state. On a long-term scalp EEG database, an average accuracy of 97%, sensitivity of 95.70%, specificity of 93.90%, and G-Mean of 94.80% were achieved, with an anticipation time of 10 min. |
Full text from SpringerLink