Development of a digital signal processing measurement platform for biomedical magnetic induction tomography and spectroscopy

Abstract

Magnetic Induction Tomography (MIT) is a non-invasive technique that utilises the passive electrical properties of a material to produce cross-sectional images. In MIT system, the signals detected by the sensors must be measured using a phase sensitive technique. A sub-millidegree phase stability is typically required for biological tissues, where the objects to be imaged have relatively low conductivity (< 2S/m). The phase noise and thermal related phase drift in the receiver's signal chain of currently available MIT systems are the major limiting factors of MIT performance for practical measurements. This thesis describes the development of a high precision DSP based signal measurement platform. It utilises multi-channel high speed digitisers to sample two or more signals simultaneously and phase differences between the signals are calculated by using FFT based algorithms. The algorithms are optimised for higher speed performance using parallel processing on both multi-core PC and graphic card processors. A faster approach based on a dedicated DSP processor for each MIT channel is later suggested to reduce data transfer speed limitations between the digitiser and the signal processing hardware. By formulating a phase noise estimation model to optimise the digitiser's setting, it is shown that better phase measurement precision and dynamic range can be achieved. To improve the phase drift for practical MIT measurements, a novel instrumentation amplifier was designed and it was incorporated into a new 5-channel annular array MIT prototype. The prototype was fully developed into a 14-channel Cardiff MIT-MKIIa system and both systems demonstrated sub-millidegree phase noise performance with a highly stable phase drift characteristic. To further investigate the MIT system for practical applications, phantom measurements were carried out to investigate the MIT system precision for detecting cerebral stroke and a single channel multi-frequency MIT system was built to perform spectroscopy measurements on biological samples.