Master thesis presentation: Improvement of an UWB time domain system for medical diagnostics

Master thesis presentation: Improvement of an UWB time domain system for medical diagnostics by Albert Monteith, MPBME.

Date: Thursday, June 11th

Time: 10.00

Place: Blå Rummet (room 6414), Hörsalsvägen 11, 6th floor

Supervisor: Xuezhi Zeng

Examiner: Andreas Fhager

Abstract
The existence of contrasts in the electrical properties of healthy tissues, cancerous tissues and blood at microwave frequencies have motivated the use of ultra-wideband (UWB) diagnostic methods. Previously developed methods for UWB microwave tomography for breast cancer detection and UWB measurements for stroke detection rely on the use of a vector network analyser (VNA). This high cost, heavy, bulky and slow instrument is not ideal for clinical use and is not suitable for use in an ambulance where this technology can have the greatest impact. An UWB time domain system has previously been developed at Chalmers University of Technology with the aim of replacing the VNA with a fast, compact and low-cost system performing UWB measurements in the time domain with sufficient accuracy. This system used short pulses to perform UWB measurements and its imaging functionality was verified by imaging tests, but there are still several problems that need to be solved in order to obtain an effective imaging system.

The aim of this project was to improve the performance of the previously developed prototype. In this work, a new clock generation system was implemented, which consequently gave a higher measurement accuracy, minimized the hardware and allowed for more flexible measurements. Both pulse trains and M-sequences were considered as stimulus signals. Simulations and measurements were carried out in order to investigate the measurement accuracy for both cases.

It was found that a signal-to-noise ratio (SNR) of at least 44 dB was achieved up to 4 GHz when the pulse train was used, which is an improvement of 8 dB at 4 GHz and up to 26 dB at lower frequencies compared to that of the previous prototype. It was also shown that the use of an M-sequence as a stimulus signal results in an SNR up to 25 dB higher than that of a pulse-train based system when the noise due to jitter is negligible. An imaging test was performed with the best system settings and the object under observation was successfully reconstructed. The complete measurement took approximately 2.5 minutes, which is approximately half the time taken for a VNA based system.