Improving the sensitivity of radar-based breast imaging algorithms in diverse patient populations
O'Loughlin, Declan Denis
MetadataShow full item record
This item's downloads: 1595 (view details)
Radar-based imaging is an emerging modality for breast cancer screening. Two commercial radar-based imaging devices are being tested in patient imaging studies. Promising initial results have highlighted the potential of the technology but have also identified that breast composition varies substantially from individual to individual. The breast composition is known to impact image quality, motivating the primary research objective: to develop novel radar-based breast imaging algorithms to address the normal variance of breast composition observed in the population. Firstly, fundamental assumptions of radar-based imaging algorithms are examined. The results of this analysis indicate that accounting for the dielectric properties of the patient-specific breast positively impacts the sensitivity of radar-based imaging in patient populations with normal breast variance. Using breast phantoms mimicking normal variation in breast composition, it is shown that using one population mean estimate of the dielectric properties may not be suitable to reconstruct images of all breasts. Secondly, a novel radar-based imaging algorithm is developed that can account for the dielectric properties of the patient-specific breast. The proposed algorithm is tested using experimental data from breast phantoms and also evaluated using five clinical case studies from the University of Calgary. The algorithm is applied to patients both with and without breast disease. These results suggest that the proposed algorithm can help improve the sensitivity in patients with breasts of differing tissue composition without impairing the specificity. However, both the experimental imaging results and clinical case studies highlight that achieving high specificity in dense breasts may be a potential challenge. More work needs to be done to investigate factors affecting the specificity of radar-based imaging. In summary, this thesis indicates that accounting for the patient-specific breast is important to achieve high sensitivity using radar-based breast imaging. A novel algorithm is proposed and tested that could be used to improve the sensitivity of radar-based breast imaging without negatively impacting the specificity.