Mohammed Rupawala

Doctoral Researcher
Physical Sciences for Health CDT

Thesis project - "Simultaneous electroencephalography and functional near-infrared spectroscopy for accurate diagnosis of prolonged disorders of consciousness"

Dr Damian Cruse, School of Psychology
Dr Hamid Dehghani, School of Computer Science
Dr Sam Lucas, School of Sport, Exercise and Rehabilitation Sciences
Professor Peter Tino, School of Computer Science

This project proposes to develop and investigate simultaneous methods for functional Near Infrared Spectroscopy (fNIRS) and electroencephalography (EEG) use to detect and diagnose prolonged disorders of consciousness (PDOC). If one is awake (except during REM-sleep) and aware of his/her environment and self, then he/she is conscious. Following severe traumatic brain injury (TBI; Glasgow Coma Scale score of 3 to 8), patients with PDOC, such as the Vegetative State (VS), do not actively respond to motor movement tasks and are therefore considered to be entirely unaware. Ethical and legal implications of withdrawing a patients life-sustaining treatment (artificial hydration andnutrition) result from such a diagnosis and so it is crucial that medical professionals are equipped with cost effective technologies that can accurately and objectively demonstrate whether a patient is conscious or not. Simultaneous EEG and functional magnetic resonance imaging (fMRI) use has shown subsets of patients in the VS to covertly follow commands to motor imagery (MI) tasks. EEG has good temporal resolution however it is a challenge to interpret the neural origins of the EEG signal due to its lack of spatial resolution, signal localisation (inverse problem), and noise contributing to the data recorded at each electrode. On the other hand, the practicality of fMRI as an accessible diagnostic tool is limited because of its cost, lack of portability, and contraindication for patients with metallic implants. Substituting fMRI for fNIRS provides high contrast images at moderate spatial resolution of the same neural response – i.e. the hemodynamic response – and has shown to detect this in ~80% of vegetative patients conducting a MI task. Nevertheless, fNIRS is limited by its poor temporal resolution due to the slow vascular response. Both fNIRS/EEG are non-invasive imaging techniques (optical fibres/electrodes on the scalp) and their simultaneous use would provide a rich dataset with improved signal sensitivity that can detect early electrical (Bereitschafts potentials) and hemodynamic responses of awareness, while maintaining portability. Combining this demonstration of motor preparation and motor activity to individual MI tasks shall provide insights into timescales for recovery and ways in which treatment can be personalised. However, there are two challenges that need to be addressed: (1) development of novel dual-modality fNIRS/EEG sensors that will allow simultaneous acquisition of data; (2) development of methods for integrating data from both modalities. Understanding the brains anatomy, neural networks, and neural origins of motor tasks is fundamental for addressing these challenges. With this knowledge, one can then process EEG/fNIRS signals (including feature extraction, data fusion and pattern recognition) before computing maps and models (e.g.general linear models, neurovascular coupling models) of healthy subjects and patients with varying degrees of consciousness. Furthermore, with robust EEG/fNIRS training datasets, one can use machine-learning methods to extract features and train models using, for example, (non) linear classification algorithms such as linear discriminant analysis, support vector machines and K-nearest neighbours, thus optimising the sensitivity and fit of model parameters to future test patient datasets. Integrating these computationally advanced analysis systems into novel simultaneous EEG/fNIRS opto-electrode sensors will have advanced physical and computational sciences, through a bedside system that better understands neuronal markers of command-following during MI tasks, and more accurately identifies patients who have been misdiagnosed as vegetative by current clinical standards, consequently improving the appropriateness of care decisions after TBI.