AIMI Research Meeting: Transforming Breast Cancer Screening with Artificial Intelligence - Helen Frazer, MD

via Zoom - email aimicenter@stanford.edu for link

Stanford community & AIMI affiliates only

Abstract:
Currently BreastScreen Australia screens 1 million well women annually for breast cancer. Screening reduces the risk of dying, however mammographic interpretation is challenging, subject to human variability, and can create harms. Despite independent double reading of all mammograms by radiologists (and a third arbitration read if disagreement) approximately 33,000 women are recalled annually for assessment and later determined not to have breast cancer (false positive), whilst approximately 1,000 women subsequently discover they have breast cancer after receiving an “all clear” result (false negative). The cost of the public breast screening program, at over $300m annually, is also rising with Australia’s ageing population. 

Recent research by the BRAIx team demonstrated opportunity to significantly improve screening outcomes, lower harms and reduce costs with AI. Our early model results, utilising global+local approaches and  developed with cancer rich screening data sets produced an AUC of 0.92. We will be launching prospective, real world, low cancer prevalent studies in July 2021. The BRAIx project is also researching novel aspects of a mammogram that may predict a woman’s risk of breast cancer. Our team has also been examining the ethical, legal and social implications of utilising AI models in healthcare, developing approaches to explain AI prediction and will use a co-creation approach to implement our models in St Vincent’s BreastScreen services which currently screen over 50,000 women annually. 

Our team brings together leading researchers from St Vincent’s Hospital Melbourne, St Vincent’s Institute, School of Population and Global Health at University of Melbourne, and Australian Institute of Machine Learning at University of Adelaide to partner with the large-scale operations of BreastScreen Victoria and St Vincent’s BreastScreen. This partnership provides access to the globally unique dataset of BreastScreen Victoria which has a digitised, cancer annotated, image dataset as well as demographic and health data, and the high-volume screening operations at St Vincent’s BreastScreen. Together the team has a unique and very realistic opportunity to quickly put into practice new AI screening models and transform the current “one size fits all” BreastScreen Australia program. 

Given the broad reach and high public awareness of BreastScreen, successful implementation will dramatically improve the AI capabilities of the interdisciplinary teams that support BreastScreen throughout Australia and increase the understanding and support from consumers and clinicians. It will establish an exemplar for broad AI deployment in healthcare and the prospect of a global software service creating value from the 30 year public investment in population health and screening data.

About:
Adjunct Associate Professor Helen Frazer is a radiologist, breast cancer clinician and AI researcher. Helen completed her MBBS in 1990, obtained her Radiology Fellowship to RANZCR in 1999 and a Masters of Epidemiology and Biostatistics in 2014. Currently Helen is Clinical Director, St Vincent’s Hospital BreastScreen, the largest breast cancer screening service in Victoria. 

Adjunct Associate Professor Frazer has over 20 years of clinical experience in breast cancer screening, imaging and diagnosis which includes working in or leading screening services across three Australian States. She brings a deep understanding of the nature of benefits, harms and challenges in population screening programmes. Her research focusses on the opportunities for technology to improve health outcomes and lower harms in cancer screening and diagnosis, and to transform the economics of screening programs. Helen has led the introduction of digital mammography, digital tomosynthesis and new digital workflows into large scaled screening services. She has significant experience translating new research and technology into practice, with attention to the users, clients and the community. Her recent research includes the use of deep learning AI for breast cancer detection with mammography; ethical, social and legal implications of AI in cancer detection; user experience and workflow studies in cancer screening; and health workforce preparation for AI tools and decisioning. She recently led the team that was awarded an Australian Government Medical Research Future Fund grant for a program to translate promising AI mammography image reading results into a new personalised cancer screening model. 

Adjunct Associate Professor Frazer invests significant time and interest in education and teaching of medical students, radiology registrars, breast imaging fellows and PhD students.