In electrical impedance tomography, a single channel failure causes distortion to the overall image. Mathematical modelling and curve-fitting techniques were used to recover corrupted data. A single channel was disconnected in two experiments on a saline-filled dish with one and two objects, respectively. Voltage gradient data were then synthesized from the overall shape of the curve and reconstructed. The technique demonstrated a considerable improvement in the image quality. We conclude that the technique can be adapted in applications where channel failure can occur regularly such as neonatal monitoring.
A medical application of a new measurement strategy developed for process tomography is presented. By exchanging the current streamlines with the equipotential lines impedance tomography can be performed using unipolar measurements. The equipotential lines are now parallel to the conducting boundary and voltage measurement is made between this boundary and an insulated electrode. A set of 32 electrodes was placed around the chest with every alternate electrode connected to form the 'conducting boundary'. A total of 120 independent measurements were available. Reciprocity error without common mode feedback was 0.5% with a dynamic range of only 1:1.12. The images were comparable to that reconstructed using adjacent drive strategy.
The feasibility of detecting the lungs in preterm babies using electrical impedance tomography (EIT) was investigated. A single frequency instrument using 16 electrodes to apply current and detect peripheral voltages was constructed. The instrument applied AC current of 1.5 mA peak-peak at a carrier frequency of 20 kHz. Images were reconstructed using a sensitivity-regions backprojection method. A 9-day-old preterm baby was tested and data were collected at a speed of 10 frames/s. A dynamic image showing the lungs at full inspiration referenced to expiration is illustrated in this paper. Impedance measurements taken across the chest during the first 2 s did not show a clear pattern thus demonstrating irregular breathing. Region of interest analysis were carried out on the reconstructed images and tracked with time. Fourier transforms were then performed on these signals and a fundamental frequency at 1 Hz, corresponding to normal breathing rate of 60 breaths/min, was detected. Harmonics of the signal caused distortion especially on the left lung where the effects of cardiac events were more dominant.
When tissue interacts with electromagnetic radiation it exhibits resistivity and permittivity changes, which decrease with frequency. Above 100 kHz it is expected that dielectric changes in tissue (permittivity) will allow one to distinguish damaged and necrotic tissue. Furthermore, tissue impedance at medium frequencies (100 kHz-1 MHz) have not been well characterized. The aim of this work was to design instrumentation for an impedance tomographic spectrometer to cover the minimum band 10 kHz-1 MHz. In order to produce images sensitive to small changes in resistivity, voltage measurement must be accurate to at least 0.1%. Using commercially available operational amplifiers, PSPICE simulations demonstrated 0.1% accuracy up to 800 kHz, falling off to 0.5% at 1 MHz. Implementation achieved a reasonably flat amplitude (+/- 0.5 dB) and a phase shift of 50 degrees from 10 kHz to 3 MHz and a receive response of 0.13 dB to 5 MHz and phase shift of -40 degrees at 3 MHz. With channel correction this design will provide useful readings up to 3 MHz.
The 'sensitivity region' method of back projection is converted from polar drive to adjacent drive current injection strategy and individual sensitivity coefficients incorporated. For each projection sensitivity coefficients for each pixel were established and pixels were grouped together into the 'sensitivity regions'. These contain all pixels that have their largest sensitivity coefficients associated with a particular measurement pair. Back projection of measurement data was performed using these coefficients as the basis of the reconstruction and less sensitive measurement pairs were ignored. Formulae for the back projection are discussed and images from computer simulated data and raw data supplied by Sheffield as part of this concerted action are presented. Comparison with polar current injection strategy is discussed.
A data interchange format is described to allow groups working on electrical impedance tomography (EIT) with disparate algorithms and instruments to compare results. The procedure has been tested by exchanging data by e-mail. The format is defined in the appendix.
Multifrequency tomography may be conveniently achieved by sequentially sweeping the probing drive current and measuring the resultant voltages at each frequency. If events change during measurement comparisons between frequencies cannot be made. Mixing several frequency components may decrease acquisition time but increase the complexity of the instrumentation. A third method is described using Fourier transformation that enables simultaneous multifrequency measurements without an increase in instrumentation. A signal is constructed from a number of sinusoidal components of known amplitude and phase. This group of components is transformed into a time series by the inverse Fourier transform and applied to the object via a voltage or current source. Transforming the detected voltage back into Fourier components will provide the frequency response of the object. Data are collected in this way for all projections and tomograms reconstructed for each frequency. This has the advantage that no special detector is required; both in-phase and quadrature components are available and systematic errors may be easily corrected by software. This technique is demonstrated using a resistor phantom with known frequency-dependent perturbations.
A two-dimensional reconstruction algorithm based on a modified version of the method of sensitivity regions is used to reconstruct data obtained from a three-dimensional finite element model. By using data obtained from off-drive-plane measurements an improved image of changes in resistivity on the drive plane is obtained.<>
A 2-d finite element model is used to simulate the adult head. The effect of gross variations of skull thickness on the sensitivity of measurement pairs to areas within the model are obtained. The implications of these measurements for electrical impedance tomography of the adult brain are discussed. Inter-patient variability of the adult skull may be a key factor in electrical impedance tomography of the adult head.
A new technique is described which exploits capacitively coupled non-electrolyte electrodes to allow two electrodes measurement of tissue section impedance. By inductively cancelling the electrode capacitance the exciting voltage is applied directly across this section. The series impedance thus measured is then only of the tissue.
Sensitivity coefficients for changes in resistivity of individual voxels within the object are dependant upon the change in resistivity within other voxels. This is demonstrated using finite element models for both on and off-plane perturbations. It is proposed that more appropriate finite element models of body sections will elicit more information for practical electrical impedance tomography reconstruction algorithms.
Pioneering work by G.S. Brindley and W.S. Lewin (1968), and W.H. Dobelle (1974) laid the foundation of technical feasible visual prosthesis. The technique is simple in concept, and image is divided into an array of small elements or pixels. The number of pixels is reduced to the number of phosphenes made available by implantation of an array of electrodes on the surface of the visual cortex. The position of these phosphenes have been previously mapped in the visual field. There are, however, a number of problems to surmount; the minimum number of electrodes required for a useable image, implant power, implant packaging and the construction of the electrodes. The present author addresses these problems and describes the prostheses being designed and constructed at the Midland Centre for Neurosurgery
