Time domain implementation of pediatric epileptic seizure detection system for enhancing the performance of detection and easy monitoring of pediatric patients. (May 2020)

• Record Type: Journal Article
• Authors: Chakrabarti, Satarupa; Swetapadma, Aleena; Ranjan, Asish; Pattnaik, Prasant Kumar
• Is Part Of: Biomedical signal processing and control. Volume 59(2020)
• Publication Date: 2020-05
• Subjects: Artificial neural network -- Discrete wavelet transform -- Pediatric seizure -- Epilepsy -- Electroencephalogram; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2020.101930
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: An improved method has been proposed for pediatric seizure detection using wavelet and ANN. A hardware set up is designed using MATLAB/SIMULINK and Arduino for easy detection and monitoring of the pediatric epileptic seizure. Proposed method focuses on pediatric epileptic seizure which is more difficult to detect unlike other methods. The highest accuracy of proposed method is found to be 95.3% for sym4 wavelet which can be implemented in real hospitals effectively. The proposed method is accurate, selective and robust which can be used in real hospitals for monitoring of the pediatric epileptic seizure patients. Abstract: Objective: The clinical phenomenon of epilepsy varies greatly among patients and this in turn, has its effect on the quality of life they lead. Studies reveal a requisite for efficient epileptic seizure detection techniques. In this work, an extensive study has been carried out for the detection of pediatric epileptic seizures. The major challenges of pediatric epileptic seizure detection lie in the extraction of appropriate features, adaptability of the method with respect to seizure signals, applicability to all types of seizure conditions and dependency on signal channels. Methods: Electroencephalogram (EEG) signals have been used as input which is processed with discrete wavelet transform (DWT) using multi-resolution analysis. Four different wavelets such as Daubechies, Symlet, Bi-orthogonal and Coiflet have been used for feature extraction. The classifier used for this work is artificial neural network (ANN). Microcontroller based prototype model has been used for substantiation of the designed architecture. Results: The work has been verified using the CHB-MIT EEG database for pediatric patients. Records of 10 patients have been selected and the measurement criteria showed the highest accuracy, sensitivity and specificity of 95.3%, 97.2% and 93.5% respectively for sym4 wavelet. Conclusion: A prototype microcontroller-based model has been designed using MATLAB/SIMULINK and ARDUINO. The advantage of the work is that it is not patient-specific or channels specific, hence it can be used to detect partial as well as generalized seizure. Significance: The significance of the proposed microcontroller-based architecture lies in its low power consumption, accurate seizure detection in a minimum amount of time and having the compatibility of working with computers and sensors. This proposed prototype can be used in the future for designing a hardware-based detection system that would be portable and non-invasive in nature.