Mammographic breast density (MBD) is a strong risk factor for developing breast cancer. MBD is typically estimated by manually selecting the area occupied by the dense tissue on a mammogram. There is interest in measuring the volume of dense tissue, or volumetric breast density (VBD), as it could potentially be a stronger risk factor. This dissertation presents and validates an algorithm to measure the VBD from digital mammograms. The algorithm is based on an empirical calibration of the mammography system, supplemented by physical modeling of x-ray imaging that includes the effects of beam polychromaticity, scattered radation, anti-scatter grid and detector glare. It also includes a method to estimate the compressed breast thickness as a function of the compression force, and a method to estimate the thickness of the breast outside of the compressed region. The algorithm was tested on 26 simulated mammograms obtained from computed tomography images, themselves deformed to mimic the effects of compression. This allowed the determination of the baseline accuracy of the algorithm. The algorithm was also used on 55 087 clinical digital mammograms, which allowed for the determination of the general characteristics of VBD and breast volume, as well as their variation as a function of age and time. The algorithm was also validated against a set of 80 magnetic resonance images, and compared against the area method on 2688 images. A preliminary study comparing association of breast cancer risk with VBD and MBD was also performed, indicating that VBD is a stronger risk factor. The algorithm was found to be accurate, generating quantitative density measurements rapidly and automatically. It can be extended to any digital mammography system, provided that the compression thickness of the breast can be determined accurately.
Identifer | oai:union.ndltd.org:TORONTO/oai:tspace.library.utoronto.ca:1807/65632 |
Date | 16 July 2014 |
Creators | Alonzo-Proulx, Olivier |
Contributors | Yaffe, Martin |
Source Sets | University of Toronto |
Language | en_ca |
Detected Language | English |
Type | Thesis |
Page generated in 0.0014 seconds