Dynamic imaging and tissue biomarker models to delineate indolent from aggressive breast calcifications

点击翻译按钮获取中文摘要

ABSTRACT. Breast cancer screening programs suffer from false positive mammograms, unnecessary biopsies, overdiagnosis, and overtreatment. A major contributor to the poor performance of screening mammography is the diagnostic and prognostic uncertainty of mammographically detected calcifications. Breast calcifications represent a biological continuum from benign disease to ductal carcinoma in situ (DCIS) to aggressive cancer. Radiologists struggle to correlate their imaging appearance with the underlying pathology and roughly two-thirds of biopsied calcifications return with a benign pathology. Although calcifications evolve dynamically over time, the current management strategy relies heavily on the static appearance of calcifications from the most recent mammogram. Most women in screening programs have multiple mammograms, yet this temporal information is consistently underutilized in clinical decision making. There is thus an urgent need to quantify the dynamics of calcifications from serial mammograms, and to characterize the relationship between calcification trajectories and disease biology. In the absence of such innovation, increasingly sensitive screening modalities are expected to further increase the burden of unnecessary diagnostic work-up and breast cancer overdiagnosis. The central hypothesis of this proposal is that dynamic imageable and tissue biomarkers contain actionable diagnostic and prognostic information about mammographic calcifications. The use of established diagnostic imaging (mammography) in conjunction with investigational imageable biomarkers will enable testing of this hypothesis. Key to this proposal will be the creation of a large database of retrospectively and prospectively collected cohorts of patients with serial mammograms, tissue samples and clinical outcomes. This proposal will consist of three specific aims: (1) Develop a static model of breast calcifications to improve the clinical performance of mammography screening; 2) Develop a dynamic model of breast calcifications to predict histopathology and DCIS prognosis; and 3) Combine the dynamic calcification model with tissue-based biomarkers of the underlying evolutionary dynamics to delineate DCIS prognosis. The proposed research is highly innovative because it adds the temporal dimension to computer-assisted classification of mammographic calcifications, yields a joint characterization of calcification growth trajectories and lesion biology, and develops dynamic risk models to predict invasive progression in women undergoing active monitoring for DCIS. This proposal will be co-led by Dr. Grimm (breast radiologist) and Dr. Ryser (mathematical modeler) supported by a highly collaborative multidisciplinary team with expertise in cancer biology, computer vision, and surgical oncology. The overall objective of this proposal is to develop a dynamic imageable biomarker that delineates lethal cancer from non- lethal disease by leveraging the temporal dimension of serial mammograms. Ultimately, the long-term goal of our work is to better identify which calcifications to biopsy (reduce unnecessary biopsies), and if pre-invasive DCIS is found, to predict whether it will remain indolent or progress to lethal cancer (reduce overtreatment).