An Integrated Deep Learning Framework for Epilepsy and Basal Ganglia Disorders Detection using Feature Extraction Technique

Abstract

Introduction: Around 1% of the world's population suffers from epilepsy. The semiology of seizures provides important clinical signs that can be used by neurologists to classify disorders. Developing automatic seizure detection techniques is needed to improve the diagnosis and monitoring of patients. Although the use of machine learning methods has been used to treat epilepsy and basal ganglia (Parkinson disease), it still requires manual adjustments for the classification procedure.

Objectives: This paper represents an integrated model combined with convolutional neural networks and long short-term memory that is designed to perform the feature extraction and classification.

Methods: The experiments were conducted to perform the binary and multi-class epilepsy normal and abnormal classes. During the process, 10 and 100 epochs count is considered during the model training and evaluation.

Results: Finally, the hyperparameter tuning helped the present work to get optimal solutions of 98.15% classification accuracy for epilepsy when compared with the other models like CNN, DNN, and 1D-CNN+LSTM.

Conclusions: The present study also worked on the feature extraction and classification process of basal ganglia disorders using a proposed 1D-CNN+LSTM model. Due to the complex data features, the proposed model is trained for 700 epochs to achieve 91% of classification result.