Breast cancer is one of the most commonly diagnosed forms of cancer amongst American women. Mammographic density has been strongly associated with breast cancer risk. Previous quantitative assessments of percent density have shown that women with higher percentage of fibroglandular tissue have a higher risk of being diagnosed with breast cancer. It has also been established that mammographic density is associated with higher water and lower fat content in the breast. This implies that women with breasts containing higher volumetric percentage of water and lower percentage of lipid are at a higher risk of developing breast cancer. The ability to quantify water and lipid in the breasts with mammographic images has the potential to stratify women based on breast cancer risk. Thus the purpose of this project was to employ dual-energy material decomposition technique to demonstrate the feasibility of quantifying water and lipid in breast tissue. Calibration study was performed to determine the fitting coefficients in the dual-energy non-linear inverse function to determine material thickness.

Postmortem breasts were imaged using a spectral mammography system and dual-energy material decomposition was implemented to quantify water and lipid in the breasts. The results from mammographic images were compared to results from chemical analysis, the reference standard. The thickness of water and lipid could be measured with RMS errors less than 1mm and density with RMS error less than 1% demonstrating a linear relationship. A high correlation between the two datasets was observed concluding that dual-energy mammography demonstrates the potential for breast cancer risk assessment based on water and lipid content.

Suspicious lesions detected during early screening requires the patient to undergo needle biopsy or further imaging examinations to determine the nature of the lesion. However, due to mammography’s low specificity for malignant lesions, several healthy women are often recalled for diagnostic procedures. Prior studies have suggested that breast lesion tissue containing higher than normal percentage of water and lower lipid have higher tendency of being malignant. The purpose of this project was also to further extend the application of dual-energy mammography to breast lesion characterization. Dual-energy images of postmortem breasts with lesion phantoms were studied. The measured thicknesses and densities of the lesions were compared to known values and a linear relationship was observed. RMS errors of less than 1mm and less than 10% were calculated for thickness and density respectively. The small errors were indicative of a good match between the measured datasets and the applied fitting function and further emphasized the accuracy of dual-energy mammography in lesion characterization.