The overall purpose of this thesis was to improve the experimental methodology of imaging brain function using EIT, with a particular focus on improving electrode application through the design of a novel servo electrode bearing helmet. Previous work, outlined in this chapter, indicates that imaging of acute stroke and focal epilepsy is feasible with EIT, but a multitude of technical limitations have precluded successful imaging from scalp recordings. The electrode helmet was designed primarily to enable imaging of acute stroke through EIT in an acute setting through rapid application of electrodes. The secondary purpose of the helmet was to address the significant problem of movement artefact and measurement drift in long term scalp recordings for the epilepsy and TBI applications of EIT. Validation of the helmet necessitated testing as representative a scenario as possible, collecting real EIT data on the scalp. The results on previous studies on the stroke ward, section 1.3.5, highlighted significant limitations in the measurement strategy, namely the SNR and modelling errors, which would mask any potential advantages conferred by the helmet design. Therefore the first two chapters of this thesis address these more fundamental limitations in EIT imaging of the head before reassessing the feasibility of imaging brain function in the subsequent chapter. With the
feasibility established, the remaining two chapters focus on the design and testing of the helmet.
In chapter 2 the issue of modelling errors in tank studies was addressed through the design of novel anatomically realistic 3D printed head tanks. Previous work by Packham, Koo, Romsauerova,et al.[68]demonstrated significant distortions in MFEIT images of a tank without a skull present, and modelling errors introduced by mismatch between tank and mesh geometry was identified as an key limitation. To address these problems novel 3D printed tanks of both the adult and neonatal head were created based on a CT and MRI segmentation, which reproduced the spatial variations of the skull conductivity. Additionally, advances in meshing and imaging by Aristovich, Santos, Packham,et al.[61]and Jehl, Dedner, Betcke, et al.[63], previously limited to simulation and recordings in the anaesthetised rat, were adapted for head tank experiments.
Chapter 3 addressed the hardware limitations identified during previous clinical and tank experiments. Drift and SNR were described as key limitations during MFEIT experiments both in tanks[68]and on the ward[7]and in TD measurements of seizures in the EEG telemetry unit[30]. Since these studies, there has been significant focus on improved EIT systems both within the group at UCL [125]and the IIRC group at Kyung Hee University, South Korea [45],[110]. However, these EIT systems had yet to be assessed with regards to brain imaging from scalp measurements, so it was not understood if the limitations identified in previous hardware had yet been overcome. To this end a comparison study was performed of the UCL ScouseTom system [125]and the KHU Mk 2.5 32 Channel system[110], in order to select the most suitable system for future experiments. The noise dependence on frequency, current amplitude and load was investigated, as well as precision and noise in tank and scalp measurements.
The work by Fabrizi, McEwan, Oh,et al.[58]demonstrated that imaging conductivity perturbations similar to that expected from seizures in a head shaped tank was possible, but with the SNR of the systems available at the time, the number of averages required made clinical recordings infeasible. The culmination of the work in the two previous chapters and the advances by Aristovich, Santos, Packham, et al. [61] and Jehl, Dedner, Betcke, et al. [63]provided an opportunity to reassess the feasibility of imaging brain function. A tank study using a similar methodology to that employed by Fabrizi, McEwan, Oh, et al. [58] was conducted in the newly constructed adult head tank, using the UCL ScouseTom system. As had been identified by Gibson, Bayford, and Holder[33], the neonatal anatomy may be more conducive to imaging with EIT, therefore a study using the same methodology was also conducted in parallel to that in the adult tank. 10 % conductivity perturbations were reconstructed in three positions inside the adult head, and five positions in the neonatal head and the images compared in terms of localisation error, resolution and image distortion. A simulation study was also performed to estimate the localisation error of perturbations
throughout the head with and without realistic noise. The final test of the feasibility was to investigate the median signal size inµV arising from these perturbations, which provided a minimum noise level which must be achieved for successful clinical images.
Finally, the final two chapters concern the main focus of the methodological improvements in this thesis: the design of the electrode bearing helmet. Despite the prevalence of abrasion during the application of electrodes for EEG and EIT, there is a paucity of objective information regarding the mechanical characteristics of the abrasion performed. Therefore, chapter five describes the investigation into the electromechanical requirements of the self abrading electrodes. To this end, the force applied during manual abrasion was characterised and a large scale test rig was designed in which the applied force, speed and torque of rotation could be independently controlled. The final chapter describes the design of a miniature self abrading electrode unit which meets the specifications outlined in chapter five, and a 32 channel electrode controller system. A prototype helmet which incorporated the self abrading electrodes was designed based on a reference head model scaled to the average head size. Finally, the helmet was then tested in scalp recordings and the extent to which the helmet met the requirements outlined in this chapter, as well as the clinical feasibility of such a system was assessed.