Semester of Graduation
Master of Science (MS)
Department of Physics and Astronomy
Breast tissue is a mixture of adipose and fibro glandular tissue with a glandular fraction on average of 20%-30%. Previous studies have confirmed that there is a risk of undetected breast cancer related to the amount of glandular tissue present in the breast. Since a higher glandular fraction can lead to an increased cancer risk, it is important for radiologists to know quantitative glandular fraction when diagnosing a patient. Another increasingly popular protocol for mammography is to eliminate the anti-scatter grid and use software algorithms to reduce scatter. This work focuses on using a maximum likelihood algorithm to estimate the pixel-wise glandular fraction (to localize the glandular tissue) from images taken with or without an anti-scatter grid. The algorithms are evaluated with Geant4-based TOPAS Monte Carlo generated images with the glandular fraction known. These images are also taken with and without microcalcifications present to study the effects of this method on microcalcification detection, as it a robust marker when analyzing a mammogram for malignant lesions. We then applied the algorithm to a few clinical DICOM images with and without microcalcifications. For the TOPAS simulated images, the glandular fraction was estimated with a root mean squared error of 3.2% and 2.5% for the without and with anti-scatter grid cases respectively along with an average absolute error of 3.7 ± 2.4% and 3.6 ± 0.9% respectively. Results from DICOM clinical images show that the algorithm gives a glandular fraction within the average range expected from the literature. This work improves on previous absolute physics models for estimating glandular fraction and shows comparable accuracy against the relative physics model.
Subsequently we investigated if adding a thickness estimation step improved the glandular fraction estimation. Here we compared a thickness estimation method (similar to the glandular fraction estimation method) and a uniform thickness method (using the paddle separation) and their effect on glandular fraction estimation. The thickness estimation yielded improved results in the standard deviation (70% lower deviation) to the uniform thickness estimation method.
Smith, Bryce, "Maximum Likelihood Estimation of Glandular Fraction for Mammography" (2023). LSU Master's Theses. 5792.
Available for download on Monday, May 18, 2026