Signal Processing and Information-Theoretic Methods for Detection and Categorisation of Cortico-Muscular Interactions

Student thesis: Doctoral ThesisDoctor of Philosophy

Abstract

This thesis aims to develop signal processing and information-theoretic methodologies for the detection and characterization of cortico-muscular in- teractions (CMI) using the electroencephalogram (EEG) recorded over the sensorimotor cortex and the electromyogram (EMG) of active muscles. Com- putational methods most commonly used in this context are cortico-muscular coherence (CMC) and Granger causality (GC) analysis, which are typically not sufficiently sensitive: some healthy subjects exhibit no significant CMC and GC, and yet have good motor skills. To enhance synchronous cortico- muscular components in mixtures captured by EEG and EMG, a concept of coherent subband independent component analysis (CoSICA) is introduced.
The methodology is accomplished through filter bank processing to decompose EEG and EMG signals into frequency bands, followed by independent com- ponent analysis, a novel component selection algorithm, and a re-synthesis of EEG and EMG to improve CMC levels. Results from simulations and neu- rophysiological signals demonstrate that CoSICA significantly enhances origi- nal CMC levels. Additionally, a multiscale wavelet transfer entropy (MWTE) methodology is proposed to develop measures of functional CMI with improved sensitivity and can detect both linear and non-linear couplings. This method- ology uses a dyadic stationary wavelet transform to decompose EEG and EMG signals into functional bands of neural oscillations and applies transfer entropy analysis with a range of embedding delay vectors to detect and quantify intra- and cross-frequency couplings at different time scales. Experiments with simu- lated and neurophysiological signals validate the potential of MWTE to detect and quantify information flows, including non-linear cross-frequency interac- tions and interactions across different temporal scales. Finally, the inferred MWTE method is applied to data collected from young people with dystonia, revealing a distinct impairment of cross-frequency feedback from muscle to brain in dystonia, indicating potential for further exploration in this field.
Date of Award1 Jun 2023
Original languageEnglish
Awarding Institution
  • King's College London
SupervisorZoran Cvetkovic (Supervisor)

Cite this

'