A new automated measure of breast cancer risk from digital and film mammography (Cirrus)

Project Details

This project has three main objectives:

  • Develop an automated algorithm for the prediction of breast cancer risk directly from film and digital presentation mammograms (Cirrus).
  • Validation of the Cirrus algorithm.
  • Deploy a proof-of-principle pilot implementation of Cirrus at BreastScreen services.

Project summary

Mammographic density is one of the strongest predictors of breast cancer risk but its impractical measurement prevents its use in a clinical setting. An automated procedure that determines which aspects of a mammogram best predict cancer would allow screening programs to identify and target women at higher risk of breast cancer, which in turn could lead to earlier diagnoses and better breast cancer outcomes. A key to this will be the actual implementation of the automated measure into real-time practice.

We will develop and validate an automated measurement, maximized by its ability to predict breast cancer risk that is directly applicable to digital presentation mammograms from a wide range of vendors. This will primarily be done by utilising the information in a digital mammogram that is unaffected by the types of processing mammography machines typically perform on their unprocessed mammograms to produce their presentation mammograms. A presentation mammogram is the image seen by radiologists during screening, and is the only mammogram routinely stored by screening services. It is therefore crucial for successful translation that an automated method should work directly on these presentation mammograms. We have already made substantial steps in this direction through an NHMRC grant (2011-2012) using analog mammograms. We now have access to large samples of digital mammograms which are essential for our method to have clinical relevance, given that digital mammography has been rolled out across BreastScreen Australia.

Researchers

Professor John Hopper

Dr Daniel Schmidt

Dr Carmel Apicella

Dr Enes Makalic

Dr Gillian Dite

Kevin Nguyen

Collaborators

Dr Helen Frazer (BreastScreen Victoria)

Dr Jennifer Stone (University of Western Australia)

Funding

NHMRC

Research Group

High Dimensional Analytics



Faculty Research Themes

Cancer

School Research Themes

Prevention and management of non-communicable diseases (including cancer), and promotion of mental health, Screening and early detection of disease



Key Contact

For further information about this research, please contact the research group leader.

Department / Centre

Centre for Epidemiology and Biostatistics