Aleksandra Vuckovic1, Vicente Jose Ferrer Gallardo2, Mohammed Jarjees3, Mathew Fraser4, Mariel Purcell4. 1. Biomedical Engineering Division, University of Glasgow, Glasgow, UK. Electronic address: Aleksandra.vuckovic@glasgow.ac.uk. 2. Biomedical Engineering Division, University of Glasgow, Glasgow, UK; Basque Center on Cognition, Brain and Language, Spain. 3. Biomedical Engineering Division, University of Glasgow, Glasgow, UK; Engineering Technical College of Mosul, The Northern Technical University, Mosul, Iraq. 4. Queen Elizabeth National Spinal Injuries Unit, Queen Elizabeth University Hospital, Glasgow, UK.
Abstract
OBJECTIVES: To create a classifier based on electroencephalography (EEG) to identify spinal cord injured (SCI) participants at risk of developing central neuropathic pain (CNP) by comparing them with patients who had already developed pain and with able bodied controls. METHODS: Multichannel EEG was recorded in the relaxed eyes opened and eyes closed states in 10 able bodied participants and 31 subacute SCI participants (11 with CNP, 10 without NP and 10 who later developed pain within 6 months of the EEG recording). Up to nine EEG band power features were classified using linear and non-linear classifiers. RESULTS: Three classifiers (artificial neural networks ANN, support vector machine SVM and linear discriminant analysis LDA) achieved similar average performances, higher than 85% on a full set of features identifying patients at risk of developing pain and achieved comparably high performance classifying between other groups. With only 10 channels, LDA and ANN achieved 86% and 83% accuracy respectively, identifying patients at risk of developing CNP. CONCLUSION: Transferable learning classifier can detect patients at risk of developing CNP. EEG markers of pain appear before its physical symptoms. Simple and complex classifiers have comparable performance. SIGNIFICANCE: Identify patients to receive prophylaxic treatment of CNP.
OBJECTIVES: To create a classifier based on electroencephalography (EEG) to identify spinal cord injured (SCI) participants at risk of developing central neuropathic pain (CNP) by comparing them with patients who had already developed pain and with able bodied controls. METHODS: Multichannel EEG was recorded in the relaxed eyes opened and eyes closed states in 10 able bodied participants and 31 subacute SCI participants (11 with CNP, 10 without NP and 10 who later developed pain within 6 months of the EEG recording). Up to nine EEG band power features were classified using linear and non-linear classifiers. RESULTS: Three classifiers (artificial neural networks ANN, support vector machine SVM and linear discriminant analysis LDA) achieved similar average performances, higher than 85% on a full set of features identifying patients at risk of developing pain and achieved comparably high performance classifying between other groups. With only 10 channels, LDA and ANN achieved 86% and 83% accuracy respectively, identifying patients at risk of developing CNP. CONCLUSION: Transferable learning classifier can detect patients at risk of developing CNP. EEG markers of pain appear before its physical symptoms. Simple and complex classifiers have comparable performance. SIGNIFICANCE: Identify patients to receive prophylaxic treatment of CNP.
Authors: Yvonne Tran; Philip Austin; Charles Lo; Ashley Craig; James W Middleton; Paul J Wrigley; Philip Siddall Journal: Sensors (Basel) Date: 2022-03-29 Impact factor: 3.576
Authors: Joshua Levitt; Muhammad M Edhi; Ryan V Thorpe; Jason W Leung; Mai Michishita; Suguru Koyama; Satoru Yoshikawa; Keith A Scarfo; Alexios G Carayannopoulos; Wendy Gu; Kyle H Srivastava; Bryan A Clark; Rosana Esteller; David A Borton; Stephanie R Jones; Carl Y Saab Journal: Neuroimage Date: 2020-08-29 Impact factor: 6.556