A Spiking Neural Network Methodology and System for Learning and Comparative Analysis of EEG Data From Healthy Versus Addiction Treated Versus Addiction Not Treated Subjects

• Published in: IEEE transactions on biomedical engineering Vol. 63; no. 9; pp. 1830 - 1841
• Main Authors: Doborjeh, Maryam Gholami, Wang, Grace Y., Kasabov, Nikola K., Kydd, Robert, Russell, Bruce
• Format: Journal Article
• Language: English
• Published: United States IEEE 01.09.2016; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Adult; Analytical models; Artificial intelligence; Biological neural networks; Brain; Brain - drug effects; Brain - physiopathology; Brain modeling; Classification; Cognition - drug effects; Cognitive tasks; Comparative analysis; Computational modeling; Data models; Diagnosis, Computer-Assisted - methods; Drug abuse; Dynamic tests; EEG data; EEG data classification; Electroencephalography; Electroencephalography (EEG) comparative analysis; Electroencephalography - methods; evolving spiking neural networks (eSNNs); Female; GO/NOGO tasks; Humans; Machine Learning; Male; Mathematical models; Methadone; Methadone - therapeutic use; methadone maintenance treatment (MMT); Narcotics; NeuCube; Neural networks; Neural Networks (Computer); Neurons; opiate addicts; Opioid-Related Disorders - drug therapy; Opioid-Related Disorders - physiopathology; Pattern Recognition, Automated - methods; Reproducibility of Results; Sensitivity and Specificity; spatiotemporal brain data (STBD)
• ISSN: 0018-9294; 1558-2531; 1558-2531
• DOI: 10.1109/TBME.2015.2503400

This paper introduces a method utilizing spiking neural networks (SNN) for learning, classification, and comparative analysis of brain data. As a case study, the method was applied to electroencephalography (EEG) data collected during a GO/NOGO cognitive task performed by untreated opiate addicts, those undergoing methadone maintenance treatment (MMT) for opiate dependence and a healthy control group. Methods: the method is based on an SNN architecture called NeuCube, trained on spatiotemporal EEG data. Objective: NeuCube was used to classify EEG data across subject groups and across GO versus NOGO trials, but also facilitated a deeper comparative analysis of the dynamic brain processes. Results: This analysis results in a better understanding of human brain functioning across subject groups when performing a cognitive task. In terms of the EEG data classification, a NeuCube model obtained better results (the maximum obtained accuracy: 90.91%) when compared with traditional statistical and artificial intelligence methods (the maximum obtained accuracy: 50.55%). Significance: more importantly, new information about the effects of MMT on cognitive brain functions is revealed through the analysis of the SNN model connectivity and its dynamics. Conclusion: this paper presented a new method for EEG data modeling and revealed new knowledge on brain functions associated with mental activity which is different from the brain activity observed in a resting state of the same subjects.