Authors:
Sourya Bhattacharyya
1
;
Jayanta Mukhopadhyay
1
;
Arun Kumar Majumdar
1
;
Bandana Majumdar
1
;
Arun Kumar Singh
2
and
Chanchal Saha
2
Affiliations:
1
Indian Institute of Technology and Kharagpur, India
;
2
SSKM Hospital, India
Keyword(s):
Electroencephalogram (EEG), Neonatal Intensive care Unit (NICU), EEG Burst Suppression, Amplitude integrated EEG (aEEG).
Related
Ontology
Subjects/Areas/Topics:
Applications and Services
;
Artificial Intelligence
;
Biomedical Engineering
;
Biomedical Signal Processing
;
Computer Vision, Visualization and Computer Graphics
;
Data Manipulation
;
Detection and Identification
;
Health Engineering and Technology Applications
;
Human-Computer Interaction
;
Medical Image Detection, Acquisition, Analysis and Processing
;
Methodologies and Methods
;
Neurocomputing
;
Neurotechnology, Electronics and Informatics
;
Pattern Recognition
;
Physiological Computing Systems
;
Sensor Networks
;
Soft Computing
Abstract:
Presence of burst suppression pattern in neonate EEG is a sign of epilepsy. Detection of burst patterns is normally done by visual inspection of recorded raw EEG or amplitude integrated EEG signal. Existing automatic burst detection approaches consist of either supervised learning mechanism or static energy threshold based comparison. Both approaches can produce inconsistent results for babies with different ages (for example, a neonate EEG and a six month old baby EEG). That is because, EEG signal amplitude or energy increases according to baby’s age. Training based classifiers or static thresholds cannot adapt with this amplitude variation. Here we propose an automatic burst detection method, which first computes signal parameters such as energy, variance and power spectral density. From generated signal data, so called low level amplitude or energy output is used as a ground reference for indication of signal suppression level. Burst is identified according to high deviation of pa
rameter values from those in suppression pattern. It does not need any static threshold based comparison. Results show that our algorithm exhibits greater sensitivity and equal specificity than existing methods. Due to adaptive thresholding for burst detection, our method is applicable for analyzing EEG signals of babies with different ages.
(More)