The full impacts of digital mammography and computer-aided diagnostic (CAD/QIA) systems on the performance of diagnostic mammography are yet to be realized. Lesion composition as described by its 3 compositional thicknesses of protein, lipid, and water (3CB) was recently discovered to be a strong descriptor of abnormal breast lesions. The long-term goal of this project is to reduce unnecessary breast biopsies by creating diagnostic imaging models using the strongest CAD/QIA algorithms incorporating advances such as 3CB. Our objective is to quantify lipid-protein-water signatures around CAD/QIA markers to better predict malignant findings. Our central hypothesis is that lesion composition can be combined with existing CAD/QIA methods to improve the specificity of cancer diagnosis and reduce the number of unnecessary biopsies.
Our specific aims are as follows: to 1) investigate the sensitivity and specificity of localized 3CB to distinguish breast cancer from benign lesions on prospectively acquired diagnostic mammograms of women recommended to undergo biopsy, 2) compare the sensitivity and specificity of 3CB to an established CAD/QIA method and conventional morphological BI-RADS descriptors, 3) develop a predictive diagnostic model to quantify the probability mammographic findings require biopsy versus don't require biopsy based on clinical risk factors, CAD/QIA measures and 3CB measures, and secondary) investigate advantages of 3-dimensional 3CB signatures using dual-energy digital breast tomosynthesis. The working hypotheses for each are as follows:
aim 1 - that unique 3CB signatures of lipid, protein, and water exist for breast cancer versus benign lesions and that this information can be used to better identify lesions that require breast biopsy aim 2 - that automated diagnostic CAD/QIA will yield quantitative lesion features that either correlate with or complement (independent) to the compositional signatures, aim 3 - that localized 3CB measures and established CAD/QIA measures are independent methods that assess different predictors of breast cancer and benign lesions and that the combination of measures from the two methods in a single model will increase the sensitivity and specificity from either one alone, secondary aim - that 3D mammography will provide more accurate lesions compositions than 2D imaging. The research's innovation is the combination of the two independent imaging risk markers: 3CB and a powerful CAD/QIA model, and compares it to the clinical standards used by radiologist. The expected outcomes include 1) a novel 3CB and CAD/QIA combined and accessible technology that will yield improved discernibility between cancerous and benign mammographic findings, 2) extensive, and biologically relevant knowledge on how lesion composition correlates with CAD/QIA findings, and 3) an demonstration of an optimized image-based predictive model for malignancy. We expected an important positive impact because more accurately identification of women with and without breast cancer will reduce the harm of unnecessary biopsies.
The proposed studies are relevant to public health because they have the potential to increase the effectiveness of diagnostic mammography. Thus, this advance is expected to have a high impact to the health and well-being of women because mammography is currently the primary diagnostic tool for early breast cancer detection, and over two-thirds of women diagnosed through mammography do not have cancer, resulting in many unnecessary painful invasive procedures. This is relevant to the part of NIH's mission which focuses on the prevention of disease by supporting research in the diagnosis of human diseases.
|Antropova, Natalia; Huynh, Benjamin Q; Giger, Maryellen L (2017) A deep feature fusion methodology for breast cancer diagnosis demonstrated on three imaging modality datasets. Med Phys 44:5162-5171|
|Brandt, Kathleen R; Scott, Christopher G; Ma, Lin et al. (2016) Comparison of Clinical and Automated Breast Density Measurements: Implications for Risk Prediction and Supplemental Screening. Radiology 279:710-9|
|Huynh, Benjamin Q; Li, Hui; Giger, Maryellen L (2016) Digital mammographic tumor classification using transfer learning from deep convolutional neural networks. J Med Imaging (Bellingham) 3:034501|
|Li, Hui; Zhu, Yitan; Burnside, Elizabeth S et al. (2016) MR Imaging Radiomics Signatures for Predicting the Risk of Breast Cancer Recurrence as Given by Research Versions of MammaPrint, Oncotype DX, and PAM50 Gene Assays. Radiology 281:382-391|
|Gierach, Gretchen L; Patel, Deesha A; Pfeiffer, Ruth M et al. (2016) Relationship of Terminal Duct Lobular Unit Involution of the Breast with Area and Volume Mammographic Densities. Cancer Prev Res (Phila) 9:149-58|
|Avila, Jesus I; Ng, Bennett K; Kotowski, Lisa et al. (2016) Can Racial Differences in Resting Metabolic Rate be Explained by Body Composition? Obesity (Silver Spring) 24:1203|
|Malkov, Serghei; Shepherd, John A; Scott, Christopher G et al. (2016) Mammographic texture and risk of breast cancer by tumor type and estrogen receptor status. Breast Cancer Res 18:122|
|Felix, Janine F; Bradfield, Jonathan P; Monnereau, Claire et al. (2016) Genome-wide association analysis identifies three new susceptibility loci for childhood body mass index. Hum Mol Genet 25:389-403|
|Felix, Ashley S; Lenz, Petra; Pfeiffer, Ruth M et al. (2016) Relationships between mammographic density, tissue microvessel density, and breast biopsy diagnosis. Breast Cancer Res 18:88|
|Oh, Hannah; Khodr, Zeina G; Sherman, Mark E et al. (2016) Relation of Serum Estrogen Metabolites with Terminal Duct Lobular Unit Involution Among Women Undergoing Diagnostic Image-Guided Breast Biopsy. Horm Cancer 7:305-315|
Showing the most recent 10 out of 27 publications