AbstractContactless electrical impedance tomography (CEIT) is a low-cost, non-invasive, non-radiation, and non-contact imaging technique which enables to image the complex permittivity distribution within the region of interest. This technique is developed based on the capacitively coupled principle, it is the combination and innovation of the traditional electrical capacitance tomography (ECT) and electrical impedance tomography (EIT). This thesis presents the experimental study on CEIT in three main application areas.
The versatile 4D skin-like capacitive imaging sensor for robotic application. With the spatial temporal total variation (ST-TV) algorithm, we used this contact less imaging technology as the robotic skin in several scenarios: pressure sensing, contactless panel which allows the real-time in air handwriting, and obstacle detection and trajectory mapping. With both the spatial and temporal as the penalty terms, ST-TV enables this artificial robotic skin to process and analyse the dynamical 3D data in real time, making this technique to be applicable in a wide range of scenarios.
The characterisation and electrical spectroscopy reconstruction of biomaterial with multi-frequency CEIT. The basement of such studies are the distinct electrical properties’ variation of organs and tissues under the different excitation frequency. The advantage of implementing CEIT for medical imaging application is that the direct contact between the electrical sensors and the conducting medium is avoided, thus avoiding the errors associated with the direct contact. In addition, CEIT shows its great potential for medical imaging use over a wide frequency range. Therefore, we were the first one to use the resistance measurements of CEIT to study the feasibility of it in breast cancer detection, with a wide frequency ranging from 200 kHz – 13 MHz. We have also extended the multi frequency CEIT technique to the small-scale level with two types of sensor arrays, providing a flexible and high temporal method for the characterisation of cells and biological tissues. In addition, we reconstructed the cole-cole diagram and complex permittivity spectroscopy for different biomaterials with multi-frequency CEIT. In these studies, the spatial-spectral total variation (SS-TV) algorithm was applied to ensure the fast convergence and well-defined reconstructed images.
The application of deep learning to CEIT. For this work, we separately developed ii the convolutional neural network (CNN) to: (i) recognise and classify the words creating by the planar CEIT sensor, (ii) reconstruct the images for two-phase material application. Especially for the CNN-based reconstructing study, the acceptable results obtaining by multi-layers CNN have confirmed its practicability and shown its own advantages over traditional inverse solving algorithms.
In summary, this thesis presents the research work of CEIT in several medical and industrial applications, these contributions can be the great basement for the future work on robotics applications, medical imaging, and intelligent medical and industrial imaging development.
|Date of Award
|1 Nov 2021
|Manuchehr Soleimani (Supervisor) & Ivan Astin (Supervisor)