Study limitations

Whilst this study represents an encouraging demonstration of the improvements in methodol- ogy within the UCL group, there are a number of limitations which prevent strong conclusions being drawn from the results. The most important element missing from this feasibility study is realistic physiological noise and measurement drift observed during scalp recordings, as described in section 1.4.2. Images were still successfully reconstructed in the simulation study with additive noise of 5µV on every channel based on scalp recordings taken with the Scouse- Tom system, which in most cases is an overestimation compared to the results in section 3.3.3. However, the multiplicative noise, in the absence of a scalp recording comparison, was based upon the percentage errors measured in the tank. Unexplained measurement drift of up to 20 % was the primary source of error found by Fabrizi, Sparkes, Horesh,et al.[30]during measurements on a telemetry ward, and until either the method is shown to be robust to those errors, or that these drifts have been minimised, it is not possible to make any claims regarding feasibility.

The SNR was sufficient to successfully image perturbations throughout both the adult and neonatal head. However the estimates were based upon averaging over 10 frames, or approximately 20 seconds. This is more than sufficient for imaging slow changes such as those arising from intra-cranial bleeding, which have a time course ranging from minutes to hours[177]. Currently, the time course of the impedance changes resulting from an epileptic activity is still unclear, and much of the early work showed impedance changes lasting the order of 10 s of seconds[84],[178]. Although, results results within the group have shown changes in impedance lasting less than a second after interictal spikes[92], time enough for only a single frame. Therefore, with the SNR of the current set up, it is unlikely that this activity would be reliably imaged.

overestimation of the amount feasible in practice, particularly in proof of principle studies using existing EEG montages. Additionally, the neonatal skull used in this study represents a somewhat idealised model from the perspective of successful imaging. The conductivity of the fontanelles is set to the same value as the brain tissue in this study, whereas in reality is likely closer to the full skull conductivity[140]. Further the fontanelles present in the model used those of a new born infant, and thus represent that largest possible aperture for current flow. As the fontanelles close, the benefits to the signal size seen in the simulation study will decrease.

4.4.6

Recommendations for future work

As the number of electrodes used in the neonatal study may not be representative of the amount feasible in initial clinical measurements, it is important to repeat the experiment with fewer electrodes in the positions commonly used in neonatal EEG, and find the optimal injection pattern for that set up. In doing so it would be possible to reduce the number of injections and thus decrease the time for each frame. Concurrent to the investigation into the minimum number of injections, experiments should be carried out to determine the sensitivity of the image reconstruction to the fontanelle size and conductivity. Construction of a model with closed fontanelles would be relatively simple with the existing CT segmentation, as would assigning a separate conductivity to the region.

The simulated voltages were generated through multiplication of the hexahedral Jacobian J_{he x} for simplicity which constitutes an “inverse crime”[170]. At the time of writing, running the forward solver for the≈ 8000 different perturbations would have been prohibitively

time consuming, but recent refinements to the PEITS solver[63], means this is now feasible. Any future feasibility studies should calculate the voltages in this manner for more accurate voltages. Much of the edge artefact which resulted in mislocation of the centre of mass with first order Tikhonov was found in the hexahedra at the very edge of the mesh. Optimisation of the hexahedral mesh to remove hexahedra on the edges of the mesh which correspond to a small number of tetrahedra in the fine mesh could thus improve image localisation.

For brevity, only the localisation error was considered in the simulation study as the correct centre of mass was deemed important for imaging epileptic foci or TBI. However, the SNR of the majority of the measurements would be less than 1, which are not likely to result in satisfactory images without data rejection. It is likely that the shape error of many of these reconstructions would be much larger than those measured in the tank in this study. Therefore consideration of the global error, and each error individually would be necessary for a complete feasibility study.

If, as described in chapter 3, the BioSemi EEG system were to be replaced with another system with a higher sampling rate, the SNR could be readily increased through the use of a higher carrier frequency. The maximum allowable current is proportional to the carrier

frequency as per IEC 60101[166], doubling the frequency would allow for twice as much current to be injected, with corresponding benefits to SNR. A higher carrier would also allow for faster switching times, and thus a higher frame rate, which could allow for imaging the faster impedance changes recorded by Vongerichten[92].

Reducing skin impedance through

abrasion: preliminary

characterisation of the effects of

rotation and force

5.1

Introduction

To enable imaging of acute stroke with Electrical Impedance Tomography an electrode-bearing helmet was proposed in chapter 1, which comprises multiple individually advancing and abrading electrodes. This chapter considers the design of a single self abrading electrode unit, built for assessment in-vitro, without the spatial limitations that would apply to a head- mounted device. This prototype was used to demonstrate the principle in a controlled test environment before testing in a more representative situation with a human test subject.

The literature regarding minimising contact impedance through abrasion, whilst clearly demonstrating the benefits is often vague as to the extent of the abrasion required. In most clinical studies, the contact impedance is measured after electrode application, and the abra- sion process is repeated until a satisfactory impedance is reached[78] [7]. Some studies have aimed at quantifying abrasion, however the results are often specific to the technique or device used, or in the case of human controlled abrasion are not transferable to other investigators. For example, a study by Tam and Webster[179]quantified the reduction of motion artefact with increasing sandpaper abrasion strokes. However, they did not quantify either the force applied during an abrasive stroke or the thickness of the skin removed per stroke. A study using a dental burr to abrade a circular area of skin showed repeatable results with the same operator, but without quantification of the torque output of the burr, nor the number of rotations or force applied it is not possible to repeat these results[180]. One proposed system punctured the skin to a repeatable depth of 0.5 mm, and 4 punctures were found to significantly reduced motion artefact[181]. However the focus of this study was not the quan-

tification of the contact impedance, and the damage to the Stratum Corneum from puncture is not readily transferred to an equivalent thickness of a layer removed. Microdermabrasion is a technique designed to remove the Stratum Corneum for cosmetic-dermatologic conditions. Studies have shown that it is possible to remove the Stratum Corneum without damage to the underlying tissues using a commercial microdermabrasion device[182]. However the problem still remains of transferring the results of this study to the acute stroke application, as the mechanical specifications and quantification of the devices are not described.

There are limitations on the design of the helmet in terms of the size and weight of the components, particularly as multiple units are required. It is important therefore that the electrode unit design be optimised to meet the mechanical requirements with a minimal size and weight. As these mechanical requirements are not described clearly in the literature, investigation was required. To achieve this, the prototype was designed to be flexible in terms of the force, torque and number of rotations applied to the surface. This enabled experiments to be performed to describe the effect of variations in force and torque on abrasion. These results will inform the design of the subsequent miniature prototype.

5.1.1

Purpose

The main purpose of the work described in this chapter was to construct a prototype self- advancing and abrading electrode, and to characterise its performance. The goal was that the electrode should meet the specifications set out in chapter 1 (1.5), by reaching an acceptable contact impedance (<5 kΩ) within 10 seconds. To achieve this, a test rig was constructed using orange skin as a test object. The target impedance was modified to 6 kΩbased on the differences between human and orange skin (section 5.2.7).

The specific questions to be answered were:

1. Does the prototype meet the specifications?

2. What are the optimum settings for:

(a) The applied force (b) The speed of rotation (c) The total angle rotated

3. What control strategy is needed? Open or closed loop, if closed which variables?

4. Is the finished design ready for use testing on human subjects? Or are there changes necessary?

5.1.2

Experimental Design

Design, validation and characterisation of prototype

Initially, a prototype self abrading electrode and test rig were constructed. The accuracy of the measurements of force, impedance and torque were evaluated using known test objects before subsequent charactisation of abrasion.. Details are given in section 5.2.7. Based on the goals set out in section 5.1.1, the criteria for determining the success of the prototype were as follows: reduce contact impedance at 20 Hz to<5 kΩin 10 seconds or less.

Characterisation of Manual Abrasion

A comparison study with the current “gold standard” abrasion technique, wherein abrasive paste in rubbed on the surface with an applicator. Contact impedance was measured before and after abrasion, and the force exerted was recorded throughout. The experimental set up is described in section 5.2.8.

Contact impedance as a function of applied force

The impedance decrease as a result of increasing pressure on the test object was measured. The measurement set-up is described in detail in section 5.2.8. The results from this experiment were used to determine the range of forces in the subsequent studies.

Impedance during abrasion over a range of applied forces

The impedance was recorded during abrasion of the test object for a range of applied forces. In each case the impedance was expressed as a function of the angle rotated by the electrode. The experiment is described in detail in section 5.2.8.

Impedance during abrasion for minimum and maximum applied torque

Abrasion was performed as with the previous experiment with the exception that the torque output of the rotation was reduced to the minimum possible value. This minimum value was determined experimentally as described in section 5.2.7.

Proof of principle on human skin

Abrasion was performed on human skin for a single applied force. The experiment was repeated with and without the application of abrasive electrode paste.