Stratified Breast Cancer Detection Based on Masking Risk in Mammography
Towards individualized breast cancer screening
About Our project
Dense tissue in the breast can hide or mask cancers in mammograms, reducing the chances for earlier detection. Using computer learning and artificial intelligence techniques we are developing a simple measure called a “masking index” that will be added to the mammography machine to automatically analyze the tissue patterns on the mammogram and inform when it is likely that density can mask the detection of a cancer. Using databases of mammograms collected at the University of Virginia and at the Sunnybrook Health Sciences Centre, where we know which cancers were detected and which could not be diagnosed using mammography, we will finalize and test the system. The knowledge gained through this research will be introduced into clinical practice, so that women with high masking scores might then be advised to have alternative screening tests that avoid the effects of masking. This is a step toward individualized breast cancer screening, ensuring that women receive the highest quality examination, while making best use of healthcare resources.
Our objectives are to:
Use Deep Learning to refine the Masking Index
Validate the Masking Index on an independent set of mammograms
Estimate the number of extra exams required in a screening population
“Shadow” implementation in a clinical environment, reader study and power determination for intervention study
This Project is a step towards more individualized breast cancer screening, ensuring that women receive the highest quality examination, while making better use of healthcare resources. At the end of this Project we will have achieved early translation to the clinic and be well along the pathway to commercialization.