PhD defence of Leonardo Gouvêa e Silva Fortaleza - Narrowband RF System with Flexible Antennas for Breast Cancer Detection
Abstract
Breast cancer is one of the leading causes of death among cancers in women. Early detection is widely considered as crucial for improving survival rates. The gold standard of mammography employs X-rays, which not only pose health risks that limit frequency of exams, but also face difficulties with dense tissues. Other clinical solutions, such as magnetic resonance imaging, have limited reach due to high costs. Microwave radar technology is a current research field for safe and effective lower-cost alternatives for breast cancer screening, relying on the dielectric contrast between healthy and tumour tissues.
Our research group investigates the use of an array with 16 flexible antennas that can be eventually implemented on a low-cost device for frequent scans at home, which would flag any suspicious anomalies for further clinical exams. This thesis focuses on design of and experiments with a narrowband system using discrete frequencies in the 2.0 – 2.2 GHz range, connected to the antenna array in a 3D printed hemisphere developed previously for an ultrawideband system that sends continuous-time pulses between 2 – 4 GHz.
Narrowband technology has the advantages of ample commercial availability and lower cost compared to ultrawideband, while the latter is capable of better temporal and spatial resolution. Traditionally, radar applications have favoured higher bandwidths, but widespread telecommunications in the microwave range operate on narrowband. This continues to elicit debate within the scientific community regarding what is more suitable for advanced applications, such as biomedical imaging.
Experimental signals with breast tissue phantoms, initially obtained in the frequency domain (FD), were analyzed to verify distinctions between tumour and baseline cases, exploring data repeatability over several measurement dates, signal losses within phantoms and the effect of noise levels, including high-powered microwave oven interference. Investigations on data variability are rare in this field, generating important results for identifying measurement uncertainty and reliability. Distinction between different phantom properties was clear from signal distributions.
Chirp Z-Transform was applied successfully to convert narrowband signals to the time domain (TD), leading to the implementation of a clutter rejection procedure (removing undesirable responses) using normalization, signal alignment and average trace subtraction, which resulted in significant improvement to signal-to-clutter ratios on a per channel basis.
Comparisons were performed between narrowband and ultrawideband system measurements, particularly in the TD, including the use of the delay-multiply-and-sum beamforming algorithm for image generation. Results expectedly provided more limited resolution, yet under certain conditions the tumour response was highlighted successfully, indicating that narrowband signals contain sufficient information for tumour identification. Also noted was that the clutter rejection techniques applied had different impacts for each system.
Additionally, a second-generation discrete-frequencies prototype concept was introduced for future work, using a software-defined radio transceiver, which will add flexibility to prototype diverse pulse bandwidths, capable of operating in either narrowband or ultrawideband. Theoretical noise analyses indicate it also provides improvements to noise performance and, thus, amplitude resolution.
Overall, the results presented here highlight the potential of narrowband technology for breast cancer detection, as well as the importance of properly processing and analyzing on the level of individual signals for extraction of relevant information.