Target-vs-One and Target-vs-All Classification of Epilepsy Using Deep Learning Technique

With the pervasive generation of medical data, there is a need for the worldwide medical and health care sector to find appropriate computational intelligence techniques for various medical conditions such as epilepsy seizures (ES). ES is a brain disorder that affects people of all ages, is a chronic, non-communicable disease, and can occur for no apparent reason owing to a genetic defect at any time. The unpredictable nature of ES poses a significant threat to human life where we have a target variable with five labels of seizure, namely pre-seizure, seizure and seizure-free, are classified using real clinical data. In order to accurately classify seizure activity (e.g., the target label) without extensive feature engineering or selection, we employ a deep learning classifier as the study’s baseline classifier. Deep learning is a branch of artificial intelligence and currently the most successful computational intelligence technique for diagnosing ES in health informatics. This paper deals with a real-life application of epilepsy classification using computational techniques namely, Target-vs-One and Target-vs-All using deep learning approach. It is investigated that the baseline classifier on Target-vs-One strategy achieved the highest f1-score and accuracy about 0.9815 and 0.9818, respectively, as compared to the performance of baseline classifier on Target-vs-All strategy (e.g., achieved 0.94 of f1-score and 0.98 of accuracy).