While mammography has been the mainstay for early detection of breast cancer for 4 decades, substantial numbers of patients with image detected breast cancer still die of the disease, underscoring that there is still room for improvin on early detection. Further challenging the current utilization of mammography as an early detection strategy is recognition that a universal screening approach leads to substantial numbers of women receiving unneeded intervention. This highlights the need to discover robust ways to identify at risk women for appropriate screening/early detection in order to maximize the benefit to harm ratio. Lastly, breast cancer is a heterogeneous disease that is captured by the molecular classification into three major clinical subtypes depending on hormonal receptors (HR: ER/PR) and human epidermal growth factor (HER2) receptor status. Understanding this heterogeneity has led to tailored treatment strategies, targeted interventions and improved outcomes for breast cancer patients. However, early detection approaches have not yet integrated this new understanding about disease biology. Therefore, in order to improve on early detection of breast cancer, efforts need to be focused on screening appropriate women- those at highest risk-, a shift to identifying earlier footprints of the malignancy [premalignant signatures], and to integrate the new understanding of molecular phenotypes into screening assessments. In this application, we will use computer- extracted mammographic features to interrogate the background breast parenchyma in order to develop subtype specific premalignant signatures for early detection. We will then determine whether molecular blood biomarkers would further refine these imaging signatures. Our hypothesis is that each breast cancer subtypes has a molecular field defect that results in unique mammographic parenchymal characteristics and that these imaging characteristics coupled with blood biomarker signatures can provide a novel, subtype specific, early detection tool.
Our aims are:1] Create a subtype specific early detection tool by integrating mammographic premalignant signatures and blood based biomarkers; 2] identify the extent of field cancerization associated with the different breast cancer subtypes and 3] Determine whether the imaging premalignant signatures precede the development of breast cancer. In a subset of women where we have multiple antecedent mammograms in our archives, we will test whether the imaging features extracted in aim #1 can be traced back to mammograms in years prior to development of cancer. If successful, the approach outlined would extend mammography from a tool that screens for the presence of cancer to a true early detection tool that identifies parenchymal hallmarks of precancerous change in subtype specific fashion. This innovative approach would thus meet the critical needs in the field of breast cancer early detection by shifting the bar for early detectio from detection of small tumors to identification of premalignant change, reducing screening for women at low risk, and integrating the heterogeneity of breast cancer within the domain of early detection.
While mammography has been the mainstay for early detection of breast cancer, substantial numbers of patients with image detected breast cancer still die of breast cancer, underscoring that there is still room for improving on 'early detection' In this application, we will use computer-extracted mammographic features and blood based biomarkers to develop subtype specific premalignant signatures for early detection. This innovative approach will address several critical needs in the field of breast cancer early detection by shifting the bar for early detection from detection of small tumors to identification f premalignant change, reducing screening for women at low risk, and integrating the heterogeneity of breast cancer within the domain of early detection.
| Li, Hui; Giger, Maryellen L; Huynh, Benjamin Q et al. (2017) Deep learning in breast cancer risk assessment: evaluation of convolutional neural networks on a clinical dataset of full-field digital mammograms. J Med Imaging (Bellingham) 4:041304 |
| Liang, Diana Hwang; El-Zein, Randa; Dave, Bhuvanesh (2015) Autophagy Inhibition to Increase Radiosensitization in Breast Cancer. J Nucl Med Radiat Ther 6: |
